# Kalshi BTC15M - Paper Trading Journal > **Simulation only.** No real orders are placed and no money is at risk. Fills are paper fills at the > displayed Kalshi top-of-book at decision time; fees follow Kalshi's published quadratic schedule. > Each 15-minute window is a binary market: YES pays $1 if BTC's close reference is at or above the > window's strike (the prior window's close reference). _Last updated: 2026-11-06T00:55:41Z - engine: `kalshi15m.engine`_ ## Desk status | | | |---|---| | Bankroll | **$971.97** (start $1,000.00, -28.03) | | Realized P/L | -28.03 over 12 trades | | Win rate | 41.7% (5W / 7L) | | Open exposure | $0.00 across 0 position(s) | | Max drawdown | $40.16 | | Windows tracked | 32 | ## Performance | Metric | Value | |---|---| | Net P/L | -28.03 | | Return on bankroll | -2.80% | | Trades settled | 12 | | Average P/L per trade | -2.34 | | Best / worst trade | +6.30 / -11.34 | | Fees paid | $7.16 | | Average entry edge | +6.8c | ### By decision maker | Source | Trades | Win rate | Net P/L | |---|---|---|---| | auto_committee | 12 | 41.7% | -28.03 | ## Trade log | Entry (UTC) | Window | Source | Side | Fill | Edge | Outcome | P/L | |---|---|---|---|---|---|---|---| | 22:52:20 | `KXBTC15M-26NOV041800` | auto_committee | **YES** | 0.28 x 53 | +13.9c | model flip: P(yes) 0.25 < 0.30 | -5.59 | | 23:46:25 | `KXBTC15M-26NOV041900` | auto_committee | **YES** | 0.33 x 45 | +4.6c | hard stop: mark 0.11 <= 0.18 | -11.34 | | 00:04:08 | `KXBTC15M-26NOV041915` | auto_committee | **NO** | 0.27 x 54 | +5.2c | model flip: P(yes) 0.71 > 0.70 | -3.06 | | 00:07:10 | `KXBTC15M-26NOV041915` | auto_committee | **NO** | 0.17 x 86 | +7.2c | model flip: P(yes) 0.83 > 0.70 | -4.98 | | 00:17:38 | `KXBTC15M-26NOV041930` | auto_committee | **YES** | 0.47 x 31 | +5.8c | hard stop: mark 0.17 <= 0.26 | -10.46 | | 01:11:24 | `KXBTC15M-26NOV042015` | auto_committee | **YES** | 0.66 x 7 | +8.5c | no | -4.73 | | 01:17:17 | `KXBTC15M-26NOV042030` | auto_committee | **NO** | 0.78 x 6 | +6.6c | no | +1.24 | | 02:01:18 | `KXBTC15M-26NOV042115` | auto_committee | **YES** | 0.41 x 11 | +6.2c | yes | +6.30 | | 22:34:22 | `KXBTC15M-26NOV051745` | auto_committee | **YES** | 0.36 x 13 | +4.2c | no | -4.89 | | 22:48:15 | `KXBTC15M-26NOV051800` | auto_committee | **NO** | 0.77 x 6 | +9.5c | no | +1.30 | | 23:47:16 | `KXBTC15M-26NOV051900` | auto_committee | **YES** | 0.60 x 8 | +5.6c | yes | +3.06 | | 00:31:16 | `KXBTC15M-26NOV051945` | auto_committee | **YES** | 0.47 x 10 | +4.9c | yes | +5.12 | ## Window-by-window ### 22:30-22:45 UTC - `KXBTC15M-26NOV041745` - settled **YES** Strike **80,580.60** - spot at open 80,808.35 - spot at close 80,827.80 - final reference 80,826.28 **No trade (fast committee).** First checkpoint at +710s: risk veto: mid 0.99 outside tradeable band [0.10, 0.90] - risk/reward too poor. 2 evaluation(s) this window; the desk stayed flat. ### 22:45-23:00 UTC - `KXBTC15M-26NOV041800` - settled **NO** Strike **80,826.28** - spot at open 80,880.18 - spot at close 80,638.95 - final reference 80,652.52 **Decision (auto_committee)** at 22:52:20 UTC: **YES** 53 contracts @ 0.28 ($14.84, fee $0.75), edge +13.9c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.37, conf 0.42): Spot down over the last minutes: 3m -0.111% (-1.58 sigma), 5m +0.021% (+0.24 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.52, conf 0.23): Spot is stretched below its 10-minute mean by -0.028% (-0.22 sigma); reversion favors YES. - **Quant desk** (Opening, P(yes)=0.33, conf 0.45): Brownian fair value: spot is -0.45 sigma from the 80826.28 strike with 7.7 min left and 0.157% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.60, conf 0.36): Order book imbalance +0.42 and 2m taker flow -0.06 point to YES pressure; spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.37, conf 0.42): Flow is not confirming the move, but price is the primary signal. Standing pat. - **Mean-reversion desk** (Rebuttal, P(yes)=0.52, conf 0.23): Price is still stretched from its mean; I keep my fade, sized small. - **Quant desk** (Rebuttal, P(yes)=0.33, conf 0.45): Model and market are broadly aligned; the edge is thin but real. - **Microstructure desk** (Rebuttal, P(yes)=0.60, conf 0.36): Flow is mixed; no strong conviction from my side. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.42, conf 0.13): P(yes)=0.42 with 67% weighted agreement and 0.13 confidence. I take YES at 0.28 x 53 contracts (14.84 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 22:45:17 | 80,880.18 | 0.51 | 0.78 | 882s | | 22:46:32 | 80,800.00 | 0.34 | 0.41 | 808s | | 22:47:13 | 80,791.80 | 0.26 | 0.36 | 766s | | 22:47:56 | 80,788.11 | 0.29 | 0.32 | 723s | | 22:48:35 | 80,825.90 | 0.43 | 0.47 | 685s | | 22:49:23 | 80,825.49 | 0.46 | 0.49 | 637s | | 22:50:06 | 80,816.41 | 0.40 | 0.45 | 594s | | 22:50:45 | 80,762.14 | 0.24 | 0.16 | 554s | | 22:51:25 | 80,751.10 | 0.18 | 0.19 | 514s | | 22:52:05 | 80,795.38 | 0.33 | 0.37 | 475s | | 22:52:43 | 80,751.50 | 0.15 | 0.21 | 436s |
**Early exit** (model flip: P(yes) 0.25 < 0.30) at 0.20 - P/L **-5.59** **LLM council review** (22:50:14 UTC) - HOLD: council pass: fee-adjusted edge +1.5c is below the 5c threshold; risk officer flags the vol input as stale after a 94-pt one-minute move; NO at 0.67 is negative EV despite being directionally right - **Bull desk**, P(yes)=0.40: Strike 80,826.28 is only 29 pts (0.036%) above spot - about a third of remaining 13-minute vol - so this is near a coin flip, well above the 0.335 mid. Brownian fair value 0.371 already beats the 0.356 fee-adjusted breakeven. Flow supports YES: 2m taker flow +0.46, book imbalance +0.06, 10m return +0.145%, spot back above its 80,791 10m mean. The 0.505->0.33 repricing plus heavy YES offer depth reads as crowded-short positioning that can squeeze. Thin but positive expected value; 55 contracts at 0.34. - **Bear desk**, P(yes)=0.38: NO is directionally favored (~62/38) but it is already paid for: NO costs 0.67 plus a 1.55c fee, roughly -8% EV. The 29-point gap is only -0.33 sigma, and the tape moved 94 pts in one minute earlier. 10m momentum is +0.145% with spot above the 10m mean, and +0.46 net taker YES buying cuts against NO. Honest answer: NO is right and not buyable here. 0 contracts; re-bid only if NO prints <= 0.61 with >5 min left. - **Risk officer**: VETO, 0 contracts. Net edge fails the 5c rule: FV 0.371 - ask 0.34 - fee 0.0157 = +1.5c; NO at 0.67 is -5.6c. The 0.155% 15m vol input is inconsistent with a 0.116% one-minute drop (~2.9 sigma), so the 0.371 fair value is a model artifact, not an edge. At a ~40% hit rate a sub-2c net edge has negative fee-adjusted expectancy. Liquidity is ample (3.2k/sec); price is the constraint. Re-quote only if YES ask <= 0.30 with internally consistent vol. - **Chair**: Bull will pay 0.34 for a 1.5c net edge; Bear agrees with the direction but refuses 0.67; Risk vetoes on the fee-adjusted edge and flags a stale vol input after a 94-point one-minute move. Collective decision: stand down. The 3.1c gross edge does not survive the 1.57c fee and the model's vol estimate is suspect. No position in KXBTC15M-26NOV041800. ### 23:00-23:15 UTC - `KXBTC15M-26NOV041815` - settled **NO** Strike **80,652.52** - spot at open 80,423.28 - spot at close 80,472.30 - final reference 80,430.04 **No trade (fast committee).** First checkpoint at +112s: risk veto: mid 0.10 outside tradeable band [0.10, 0.90] - risk/reward too poor. 11 evaluation(s) this window; the desk stayed flat. ### 23:15-23:30 UTC - `KXBTC15M-26NOV041830` - settled **NO** Strike **80,430.04** - spot at open 80,441.30 - spot at close 80,424.23 - final reference 80,428.06 **No trade (fast committee).** First checkpoint at +365s: best net edge -0.002 is below the 0.04 threshold (P(yes)=0.39, yes 0.41/0.42). 8 evaluation(s) this window; the desk stayed flat. **LLM council review** (23:19:47 UTC) - HOLD: council pass: 3.3c net edge is below the 4c gate; the model's drift term accounted for 0.165 sigma of the 0.277-sigma distance, so the vol-consistent fair value is ~0.55 vs a 0.56 ask; flow is offer-heavy and taker flow is selling YES - **Bull desk**, P(yes)=0.60: Spot 80,457.64 is +0.28 sigma over the 80,430.04 strike with 13.8 min left; Brownian FV 0.609 vs 0.56 ask is 4.9c gross, 3.3c net. The 10-minute mean is stale - the last 5 minutes added +0.121% and held. Offer-heavy book lets us buy a favorable settlement probability cheap; buyers defended 0.555. Buy 40 at 0.56 (~$22.40). Price stop: exit if YES bid <= 0.45. Time stop: flatten by 23:28:30Z. Trim half at 0.66. - **Bear desk**, P(yes)=0.57: Fade the pop, but it is not buyable. +0.034% is 27 points - inside 13.8 minutes of 0.3% noise. The 10-minute mean (80,402.83) is below strike; price crossed only minutes ago. YES spiked 0.535->0.595 in 30 seconds on that thin pop and has already bled to 0.56; 1m -0.010% and 10m -0.054% show momentum stalling. Offer-heavy book (-0.15) and -0.50 net taker YES selling. NO at 0.45 needs P(NO) > 0.467 post-fee; fair is 0.39-0.46. Zero contracts. - **Risk officer**: VETO, 0 contracts. Net edge 3.3c is below the 4c minimum. Fair value is suspect: 0.609 implies ~0.12% vol against 0.318% realized; at realized vol P(YES) is ~0.55, so the 0.56 ask is flat to negative. Flow opposes: book -0.15, -0.50 taker selling, 10m mean below strike. Elevated vol right after a momentum loss is the wrong time for a marginal bet. Require >=4c net with flow confirmation and realized vol back under ~0.25%. - **Chair**: Unanimous pass. The Risk officer's vol objection is the decisive one: recomputing the model without the extrapolated 5-minute momentum term gives P(YES) ~0.55, which makes the 0.56 ask a 1c loser after fees. The 0.609 fair value was 0.165 sigma of drift, not diffusion - momentum was being counted twice, once in the drift term and again in the momentum desk's tilt. No position in KXBTC15M-26NOV041830. ### 23:30-23:45 UTC - `KXBTC15M-26NOV041845` - settled **YES** Strike **80,428.06** - spot at open 80,406.10 - spot at close 80,632.00 - final reference 80,625.72 **No trade (fast committee).** First checkpoint at +65s: best net edge -0.012 is below the 0.04 threshold (P(yes)=0.53, yes 0.51/0.52). 12 evaluation(s) this window; the desk stayed flat. ### 23:45-00:00 UTC - `KXBTC15M-26NOV041900` - settled **NO** Strike **80,625.72** - spot at open 80,614.90 - spot at close 80,428.31 - final reference 80,419.69 **Decision (auto_committee)** at 23:46:25 UTC: **YES** 45 contracts @ 0.33 ($14.85, fee $0.70), edge +4.6c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.42, conf 0.36): Spot down over the last minutes: 3m -0.071% (-0.87 sigma), 5m -0.016% (-0.15 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.52, conf 0.22): Spot is stretched below its 10-minute mean by -0.024% (-0.16 sigma); reversion favors YES. - **Quant desk** (Opening, P(yes)=0.38, conf 0.43): Brownian fair value: spot is -0.30 sigma from the 80625.72 strike with 13.7 min left and 0.184% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.61, conf 0.45): Book is offer-heavy (imbalance -0.14) and the 2-minute tape shows net taker buying of YES (+0.64); spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.42, conf 0.36): Flow is not confirming the move, but price is the primary signal. Standing pat. - **Mean-reversion desk** (Rebuttal, P(yes)=0.52, conf 0.22): Price is still stretched from its mean; I keep my fade, sized small. - **Quant desk** (Rebuttal, P(yes)=0.38, conf 0.43): Model and market are broadly aligned; the edge is thin but real. - **Microstructure desk** (Rebuttal, P(yes)=0.61, conf 0.45): Flow is mixed; no strong conviction from my side. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.39, conf 0.06): Quant anchor P(yes)=0.38, tilted +0.005 by the discretionary desks, shrunk to 0.39 for model error; 62% weighted agreement, 0.06 confidence. I take YES at 0.33 x 45 contracts (14.85 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 23:45:43 | 80,614.90 | 0.43 | 0.50 | 857s | | 23:46:42 | 80,593.90 | 0.34 | 0.40 | 797s | | 23:47:52 | 80,698.00 | 0.48 | 0.69 | 727s | | 23:48:49 | 80,643.83 | 0.42 | 0.54 | 670s | | 23:49:51 | 80,592.00 | 0.29 | 0.41 | 609s | | 23:50:40 | 80,569.00 | 0.23 | 0.35 | 560s | | 23:51:28 | 80,493.70 | 0.07 | 0.19 | 511s | | 23:52:20 | 80,494.60 | 0.06 | 0.19 | 459s | | 23:53:19 | 80,556.00 | 0.05 | 0.30 | 401s | | 23:54:18 | 80,479.89 | 0.02 | 0.13 | 342s | | 23:55:09 | 80,477.60 | 0.02 | 0.12 | 291s |
**Early exit** (hard stop: mark 0.11 <= 0.18) at 0.10 - P/L **-11.34** ### 00:00-00:15 UTC - `KXBTC15M-26NOV041915` - settled **YES** Strike **80,419.69** - spot at open 80,473.50 - spot at close 80,662.01 - final reference 80,631.87 **Decision (auto_committee)** at 00:04:08 UTC: **NO** 54 contracts @ 0.27 ($14.58, fee $0.75), edge +5.2c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.51, conf 0.26): Spot up over the last minutes: 3m -0.029% (-0.31 sigma), 5m +0.088% (+0.73 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.49, conf 0.22): Spot is stretched above its 10-minute mean by +0.020% (+0.12 sigma); reversion favors NO. - **Quant desk** (Opening, P(yes)=0.67, conf 0.44): Brownian fair value: spot is +0.44 sigma from the 80419.69 strike with 10.9 min left and 0.210% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.54, conf 0.26): Book is bid-heavy (imbalance +0.08) and the 2-minute tape shows net taker buying of YES (+0.06); spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.51, conf 0.27): Order flow points the same way as price. I hold my conviction. - **Mean-reversion desk** (Rebuttal, P(yes)=0.49, conf 0.22): Price is still stretched from its mean; I keep my fade, sized small. - **Quant desk** (Rebuttal, P(yes)=0.67, conf 0.44): Model and market are broadly aligned; the edge is thin but real. - **Microstructure desk** (Rebuttal, P(yes)=0.54, conf 0.26): Flow is mixed; no strong conviction from my side. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.66, conf 0.19): Quant anchor P(yes)=0.67, tilted +0.003 by the discretionary desks, shrunk to 0.66 for model error; 90% weighted agreement, 0.19 confidence. I take NO at 0.27 x 54 contracts (14.58 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 00:03:02 | 80,473.50 | 0.69 | 0.64 | 717s | | 00:03:38 | 80,476.00 | 0.68 | 0.66 | 681s | | 00:04:19 | 80,473.60 | 0.69 | 0.67 | 641s | | 00:04:56 | 80,514.59 | 0.78 | 0.77 | 604s | | 00:05:35 | 80,528.10 | 0.89 | 0.80 | 565s | | 00:06:11 | 80,529.11 | 0.86 | 0.80 | 529s | | 00:06:48 | 80,495.70 | 0.81 | 0.72 | 491s | | 00:07:24 | 80,505.40 | 0.83 | 0.78 | 455s | | 00:08:00 | 80,473.40 | 0.72 | 0.68 | 419s | | 00:08:36 | 80,500.19 | 0.84 | 0.79 | 383s | | 00:09:12 | 80,498.70 | 0.88 | 0.80 | 347s |
**Early exit** (model flip: P(yes) 0.71 > 0.70) at 0.24 - P/L **-3.06** **Decision (auto_committee)** at 00:07:10 UTC: **NO** 86 contracts @ 0.17 ($14.62, fee $0.85), edge +7.2c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.53, conf 0.28): Spot up over the last minutes: 3m +0.026% (+0.27 sigma), 5m +0.006% (+0.05 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.48, conf 0.23): Spot is stretched above its 10-minute mean by +0.037% (+0.21 sigma); reversion favors NO. - **Quant desk** (Opening, P(yes)=0.75, conf 0.47): Brownian fair value: spot is +0.67 sigma from the 80419.69 strike with 8.0 min left and 0.214% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.63, conf 0.38): Book is bid-heavy (imbalance +0.25) and the 2-minute tape shows net taker buying of YES (+0.27); spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.53, conf 0.30): Order flow points the same way as price. I hold my conviction. - **Mean-reversion desk** (Rebuttal, P(yes)=0.48, conf 0.23): Price is still stretched from its mean; I keep my fade, sized small. - **Quant desk** (Rebuttal, P(yes)=0.75, conf 0.47): Model and market are broadly aligned; the edge is thin but real. - **Microstructure desk** (Rebuttal, P(yes)=0.63, conf 0.38): Flow is mixed; no strong conviction from my side. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.75, conf 0.30): Quant anchor P(yes)=0.75, tilted +0.013 by the discretionary desks, shrunk to 0.75 for model error; 91% weighted agreement, 0.30 confidence. I take NO at 0.17 x 86 contracts (14.62 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 00:03:02 | 80,473.50 | 0.69 | 0.64 | 717s | | 00:03:38 | 80,476.00 | 0.68 | 0.66 | 681s | | 00:04:19 | 80,473.60 | 0.69 | 0.67 | 641s | | 00:04:56 | 80,514.59 | 0.78 | 0.77 | 604s | | 00:05:35 | 80,528.10 | 0.89 | 0.80 | 565s | | 00:06:11 | 80,529.11 | 0.86 | 0.80 | 529s | | 00:06:48 | 80,495.70 | 0.81 | 0.72 | 491s | | 00:07:24 | 80,505.40 | 0.83 | 0.78 | 455s | | 00:08:00 | 80,473.40 | 0.72 | 0.68 | 419s | | 00:08:36 | 80,500.19 | 0.84 | 0.79 | 383s | | 00:09:12 | 80,498.70 | 0.88 | 0.80 | 347s |
**Early exit** (model flip: P(yes) 0.83 > 0.70) at 0.13 - P/L **-4.98** ### 00:15-00:30 UTC - `KXBTC15M-26NOV041930` - settled **NO** Strike **80,631.87** - spot at open 80,662.01 - spot at close 80,669.51 - final reference 80,629.98 **Decision (auto_committee)** at 00:17:38 UTC: **YES** 31 contracts @ 0.47 ($14.57, fee $0.55), edge +5.8c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.49, conf 0.26): Spot down over the last minutes: 3m -0.006% (-0.07 sigma), 5m -0.006% (-0.05 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.42, conf 0.31): Spot is stretched above its 10-minute mean by +0.127% (+0.72 sigma); reversion favors NO. - **Quant desk** (Opening, P(yes)=0.56, conf 0.41): Brownian fair value: spot is +0.15 sigma from the 80631.87 strike with 12.5 min left and 0.216% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.43, conf 0.34): Book is offer-heavy (imbalance -0.32) and the 2-minute tape shows net taker buying of YES (+0.07); spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.49, conf 0.21): The quant model sits at P=0.56, against my trend read. Two different lenses, so I cut conviction rather than fight it. - **Mean-reversion desk** (Rebuttal, P(yes)=0.42, conf 0.31): Price is still stretched from its mean; I keep my fade, sized small. - **Quant desk** (Rebuttal, P(yes)=0.56, conf 0.41): Model and market are broadly aligned; the edge is thin but real. - **Microstructure desk** (Rebuttal, P(yes)=0.43, conf 0.34): Flow is mixed; no strong conviction from my side. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.55, conf 0.00): Quant anchor P(yes)=0.56, tilted -0.013 by the discretionary desks, shrunk to 0.55 for model error; 51% weighted agreement, 0.00 confidence. I take YES at 0.47 x 31 contracts (14.57 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 00:15:31 | 80,662.01 | 0.51 | 0.62 | 868s | | 00:17:28 | 80,656.90 | 0.46 | 0.56 | 751s | | 00:18:44 | 80,608.20 | 0.36 | 0.42 | 675s | | 00:19:53 | 80,649.99 | 0.59 | 0.55 | 607s | | 00:20:57 | 80,647.30 | 0.51 | 0.55 | 543s | | 00:22:17 | 80,650.90 | 0.79 | 0.58 | 463s | | 00:24:15 | 80,681.60 | 0.54 | 0.75 | 345s | | 00:26:46 | 80,637.10 | 0.52 | 0.53 | 193s | | 00:27:49 | 80,577.70 | 0.07 | 0.14 | 131s | | 00:28:50 | 80,601.05 | 0.08 | 0.18 | 70s | | 00:29:52 | 80,639.30 | 0.03 | 0.68 | 7s |
**Early exit** (hard stop: mark 0.17 <= 0.26) at 0.16 - P/L **-10.46** ### 00:30-00:45 UTC - `KXBTC15M-26NOV041945` - settled **YES** Strike **80,629.98** - spot at open 80,657.72 - spot at close 80,717.48 - final reference 80,678.64 **No trade (fast committee).** First checkpoint at +75s: best net edge -0.003 is below the 0.04 threshold (P(yes)=0.67, yes 0.65/0.66). 12 evaluation(s) this window; the desk stayed flat. ### 00:45-01:00 UTC - `KXBTC15M-26NOV042000` - settled **NO** Strike **80,678.64** - spot at open 80,740.80 - spot at close 80,474.61 - final reference 80,469.66 **No trade (fast committee).** First checkpoint at +68s: best net edge +0.005 is below the 0.04 threshold (P(yes)=0.72, yes 0.69/0.70). 13 evaluation(s) this window; the desk stayed flat. **LLM council review** (00:48:16 UTC) - HOLD: unanimous pass: 1.6c net edge is below the 4c gate, book imbalance -0.30 and taker flow -0.36 oppose YES; council endorses hold-to-settlement but requires stake cut from 1.0% to 0.5% and a larger out-of-sample backtest before the policy is locked in - **Bull desk**, P(yes)=0.70: Fair 0.703 vs ask 0.67 is only ~1.6c net of fees, under the 4c gate, and book imbalance -0.30 with 2m taker flow -0.36 both lean against YES. Thin edge plus adverse flow: pass. On policy: endorse hold-to-settlement, +$702 vs +$60 is decisive and stops likely hurt a 41.6%-win payoff. But $383 drawdown is roughly 25% of a 1%-sized book: cut to 0.5% and verify there is no look-ahead in the backtest. - **Bear desk**, P(yes)=0.67: No NO trade: NO at 0.34 is ~4c rich versus 0.297 fair. Bearish flow (-0.30 book, -0.36 taker) and flat 1m argue caution, but +0.53 sigma and positive 3-10m momentum keep YES favored; the edge sits on YES, not NO. On policy: endorse settlement holding at 1% stake. Stops captured +$60 versus +$702 held; at a 41.6% win rate, stopping out on noise is the leak. $383 drawdown is tolerable at 1% sizing, but pre-commit to that cap and do not scale up. - **Risk officer**: VETO this window: net edge 1.6c after the 7% fee against a 4c minimum; imbalance -0.30 and taker flow -0.36 contradict the +0.53 sigma; no trade. On the policy: reject it as an unconditional rule. A $383 drawdown on $1,000 is 38% and blows through both the 1% risk budget and the 12% daily breaker, and 267 trades is too thin to justify deleting every exit. Holding may beat churn, but only with per-window edge gates, sizing that caps worst-case settlement loss at 1%, and a hard daily breaker. Re-test with 1,000+ out-of-sample windows before adopting. - **Chair**: Unanimous no-trade: the fee-adjusted edge is 1.6c against a 4c gate and both flow measures oppose YES, so KXBTC15M-26NOV042000 is passed. The council accepts the backtest finding that early exits were the leak (the +$702 vs +$60 comparison, with the look-ahead bug already found and fixed), but the risk officer's sizing objection carries: a 38% drawdown is not acceptable at 1% per trade. Action: stake halved to 0.5% of bankroll, which caps worst-case settlement loss at ~0.5% per window, and the sample must be extended beyond 400 windows before the policy is treated as settled. ### 01:00-01:15 UTC - `KXBTC15M-26NOV042015` - settled **NO** Strike **80,469.66** - spot at open 80,468.84 - spot at close 80,453.07 - final reference 80,464.18 **Decision (auto_committee)** at 01:11:24 UTC: **YES** 7 contracts @ 0.66 ($4.62, fee $0.11), edge +8.5c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.60, conf 0.38): Spot up over the last minutes: 3m +0.085% (+0.92 sigma), 5m +0.041% (+0.34 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.45, conf 0.26): Spot is stretched above its 10-minute mean by +0.069% (+0.41 sigma); reversion favors NO. - **Quant desk** (Opening, P(yes)=0.76, conf 0.47): Brownian fair value: spot is +0.72 sigma from the 80469.66 strike with 3.7 min left and 0.207% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.56, conf 0.32): Book is offer-heavy (imbalance -0.03) and the 2-minute tape shows net taker buying of YES (+0.32); spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.60, conf 0.40): Order flow points the same way as price. I hold my conviction. - **Mean-reversion desk** (Rebuttal, P(yes)=0.45, conf 0.26): Price is still stretched from its mean; I keep my fade, sized small. - **Quant desk** (Rebuttal, P(yes)=0.76, conf 0.47): Model and market are broadly aligned; the edge is thin but real. - **Microstructure desk** (Rebuttal, P(yes)=0.56, conf 0.32): Flow is mixed; no strong conviction from my side. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.76, conf 0.29): Quant anchor P(yes)=0.76, tilted +0.011 by the discretionary desks, shrunk to 0.76 for model error; 90% weighted agreement, 0.30 confidence. I take YES at 0.66 x 7 contracts (4.62 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 01:00:24 | 80,468.84 | 0.49 | 0.43 | 876s | | 01:01:52 | 80,425.62 | 0.38 | 0.34 | 787s | | 01:03:20 | 80,470.83 | 0.47 | 0.56 | 699s | | 01:04:46 | 80,441.24 | 0.40 | 0.41 | 613s | | 01:06:20 | 80,502.34 | 0.56 | 0.60 | 519s | | 01:07:47 | 80,482.20 | 0.53 | 0.54 | 432s | | 01:09:12 | 80,446.12 | 0.28 | 0.41 | 348s | | 01:10:39 | 80,515.99 | 0.65 | 0.70 | 260s | | 01:12:07 | 80,495.67 | 0.24 | 0.66 | 173s | | 01:13:35 | 80,475.04 | 0.12 | 0.29 | 84s |
**Result:** NO - P/L **-4.73** - bankroll after $971.97 ### 01:15-01:30 UTC - `KXBTC15M-26NOV042030` - settled **NO** Strike **80,464.18** - spot at open 80,407.50 - spot at close 80,320.38 - final reference 80,327.61 **Decision (auto_committee)** at 01:17:17 UTC: **NO** 6 contracts @ 0.78 ($4.68, fee $0.08), edge +6.6c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.20, conf 0.70): Spot down over the last minutes: 3m -0.223% (-2.70 sigma), 5m -0.231% (-2.16 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.61, conf 0.35): Spot is stretched below its 10-minute mean by -0.154% (-1.02 sigma); reversion favors YES. - **Quant desk** (Opening, P(yes)=0.13, conf 0.51): Brownian fair value: spot is -1.14 sigma from the 80464.18 strike with 12.9 min left and 0.185% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.63, conf 0.40): Book is bid-heavy (imbalance +0.03) and the 2-minute tape shows net taker buying of YES (+0.57); spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.20, conf 0.70): Flow is not confirming the move, but price is the primary signal. Standing pat. - **Mean-reversion desk** (Rebuttal, P(yes)=0.61, conf 0.21): Trend score -2.51 with flow confirming. Fading a live trend is a low-probability game - I stand down. - **Quant desk** (Rebuttal, P(yes)=0.13, conf 0.51): Model and market are broadly aligned; the edge is thin but real. - **Microstructure desk** (Rebuttal, P(yes)=0.63, conf 0.40): Flow is mixed; no strong conviction from my side. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.14, conf 0.36): Quant anchor P(yes)=0.13, tilted -0.004 by the discretionary desks, shrunk to 0.14 for model error; 72% weighted agreement, 0.36 confidence. I take NO at 0.78 x 6 contracts (4.68 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 01:15:25 | 80,407.50 | 0.30 | 0.40 | 874s | | 01:16:53 | 80,337.23 | 0.23 | 0.16 | 786s | | 01:18:21 | 80,335.75 | 0.21 | 0.15 | 698s | | 01:19:48 | 80,234.11 | 0.08 | 0.04 | 611s | | 01:21:13 | 80,203.53 | 0.06 | 0.02 | 527s | | 01:22:38 | 80,218.58 | 0.04 | 0.02 | 442s | | 01:24:03 | 80,324.70 | 0.10 | 0.11 | 356s | | 01:25:29 | 80,388.59 | 0.23 | 0.25 | 271s | | 01:26:55 | 80,373.73 | 0.08 | 0.15 | 185s | | 01:28:21 | 80,346.79 | 0.01 | 0.02 | 98s | | 01:29:52 | 80,314.05 | 0.01 | 0.00 | 7s |
**Result:** NO - P/L **+1.24** - bankroll after $971.97 ### 01:30-01:45 UTC - `KXBTC15M-26NOV042045` - settled **YES** Strike **80,327.61** - spot at open 80,341.64 - spot at close 80,650.68 - final reference 80,630.00 **No trade (fast committee).** First checkpoint at +70s: best net edge +0.016 is below the 0.04 threshold (P(yes)=0.61, yes 0.64/0.65). 12 evaluation(s) this window; the desk stayed flat. ### 01:45-02:00 UTC - `KXBTC15M-26NOV042100` - settled **YES** Strike **80,630.00** - spot at open 80,730.76 - spot at close 80,613.61 - final reference 80,635.04 **No trade (fast committee).** First checkpoint at +73s: best net edge +0.006 is below the 0.04 threshold (P(yes)=0.67, yes 0.69/0.70). 12 evaluation(s) this window; the desk stayed flat. ### 02:00-02:15 UTC - `KXBTC15M-26NOV042115` - settled **YES** Strike **0.00** - spot at open 80,597.26 - spot at close 80,745.12 - final reference 80,660.48 **Decision (auto_committee)** at 02:01:18 UTC: **YES** 11 contracts @ 0.41 ($4.51, fee $0.19), edge +6.2c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.44, conf 0.33): Spot down over the last minutes: 3m -0.042% (-0.46 sigma), 5m -0.049% (-0.42 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.55, conf 0.27): Spot is stretched below its 10-minute mean by -0.079% (-0.48 sigma); reversion favors YES. - **Quant desk** (Opening, P(yes)=0.49, conf 0.40): Brownian fair value: spot is -0.03 sigma from the 0.00 strike with 13.7 min left and 0.204% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.53, conf 0.45): Book is offer-heavy (imbalance -0.28) and the 2-minute tape shows net taker buying of YES (+0.49); spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.44, conf 0.33): Flow is not confirming the move, but price is the primary signal. Standing pat. - **Mean-reversion desk** (Rebuttal, P(yes)=0.55, conf 0.27): Price is still stretched from its mean; I keep my fade, sized small. - **Quant desk** (Rebuttal, P(yes)=0.49, conf 0.40): Model and market are broadly aligned; the edge is thin but real. - **Microstructure desk** (Rebuttal, P(yes)=0.53, conf 0.45): Flow is mixed; no strong conviction from my side. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.49, conf 0.00): Quant anchor P(yes)=0.49, tilted +0.002 by the discretionary desks, shrunk to 0.49 for model error; 60% weighted agreement, 0.00 confidence. I take YES at 0.41 x 11 contracts (4.51 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 02:00:15 | 80,597.26 | 0.41 | 0.44 | 885s | | 02:00:52 | 80,604.39 | 0.43 | 0.47 | 847s | | 02:01:28 | 80,616.72 | 0.46 | 0.48 | 811s | | 02:02:04 | 80,611.17 | 0.43 | 0.49 | 775s | | 02:02:40 | 80,598.18 | 0.39 | 0.48 | 739s | | 02:03:22 | 80,625.15 | 0.46 | 0.49 | 698s | | 02:04:02 | 80,677.99 | 0.64 | 0.51 | 658s | | 02:04:39 | 80,714.08 | 0.76 | 0.53 | 621s | | 02:05:16 | 80,733.03 | 0.77 | 0.54 | 584s | | 02:05:52 | 80,779.03 | 0.85 | 0.55 | 547s | | 02:06:28 | 80,729.71 | 0.74 | 0.53 | 511s |
**Result:** YES - P/L **+6.30** - bankroll after $971.97 ### 07:00-07:15 UTC - `KXBTC15M-26NOV050215` - settled **YES** Strike **80,553.75** - spot at open 0.00 - spot at close 80,745.12 - final reference 80,575.33 _No decision recorded._ ### 10:15-10:30 UTC - `KXBTC15M-26NOV050530` - settled **YES** Strike **80,323.83** - spot at open 0.00 - spot at close 80,745.12 - final reference 80,419.18 _No decision recorded._ ### 10:30-10:45 UTC - `KXBTC15M-26NOV050545` - settled **NO** Strike **80,419.18** - spot at open 0.00 - spot at close 80,745.12 - final reference 80,254.55 _No decision recorded._ ### 10:45-11:00 UTC - `KXBTC15M-26NOV050600` - settled **NO** Strike **80,254.55** - spot at open 0.00 - spot at close 80,745.12 - final reference 80,185.06 _No decision recorded._ ### 17:30-17:45 UTC - `KXBTC15M-26NOV051245` - settled **YES** Strike **75,766.97** - spot at open 0.00 - spot at close 80,745.12 - final reference 75,859.45 _No decision recorded._ ### 22:00-22:15 UTC - `KXBTC15M-26NOV051715` - settled **YES** Strike **75,725.90** - spot at open 75,929.27 - spot at close 75,971.40 - final reference 75,965.76 _No decision recorded._ ### 22:15-22:30 UTC - `KXBTC15M-26NOV051730` - settled **NO** Strike **75,965.76** - spot at open 75,978.04 - spot at close 75,866.40 - final reference 75,870.02 **No trade (fast committee).** First checkpoint at +73s: best net edge +0.014 is below the 0.04 threshold (P(yes)=0.48, yes 0.44/0.45). 7 evaluation(s) this window; the desk stayed flat. ### 22:30-22:45 UTC - `KXBTC15M-26NOV051745` - settled **NO** Strike **75,870.02** - spot at open 75,827.74 - spot at close 75,704.07 - final reference 75,723.21 **Decision (auto_committee)** at 22:34:22 UTC: **YES** 13 contracts @ 0.36 ($4.68, fee $0.21), edge +4.2c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.43, conf 0.34): Spot down over the last minutes: 3m -0.079% (-0.74 sigma), 5m -0.008% (-0.06 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.51, conf 0.21): Spot is stretched below its 10-minute mean by -0.017% (-0.09 sigma); reversion favors YES. - **Quant desk** (Opening, P(yes)=0.39, conf 0.43): Brownian fair value: spot is -0.27 sigma from the 75870.02 strike with 10.6 min left and 0.239% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.77, conf 0.60): Book is bid-heavy (imbalance +0.28) and the 2-minute tape shows net taker buying of YES (+0.97); spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.43, conf 0.34): Flow is not confirming the move, but price is the primary signal. Standing pat. - **Mean-reversion desk** (Rebuttal, P(yes)=0.51, conf 0.21): Price is still stretched from its mean; I keep my fade, sized small. - **Quant desk** (Rebuttal, P(yes)=0.39, conf 0.43): Model and market are broadly aligned; the edge is thin but real. - **Microstructure desk** (Rebuttal, P(yes)=0.77, conf 0.63): Book and tape agree with each other, which is rare enough to trust. Keeping size. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.42, conf 0.06): Quant anchor P(yes)=0.39, tilted +0.022 by the discretionary desks, shrunk to 0.42 for model error; 44% weighted agreement, 0.06 confidence. I take YES at 0.36 x 13 contracts (4.68 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 22:34:21 | 75,827.74 | 0.35 | 0.39 | 639s | | 22:35:57 | 75,864.39 | 0.54 | 0.48 | 542s | | 22:37:24 | 75,820.35 | 0.27 | 0.31 | 456s | | 22:38:12 | 75,782.93 | 0.21 | 0.16 | 408s | | 22:39:00 | 75,774.88 | 0.17 | 0.10 | 360s | | 22:39:48 | 75,660.61 | 0.04 | 0.01 | 312s | | 22:41:12 | 75,740.40 | 0.11 | 0.04 | 228s | | 22:42:00 | 75,739.09 | 0.04 | 0.01 | 180s | | 22:42:48 | 75,756.86 | 0.03 | 0.03 | 132s | | 22:43:36 | 75,776.32 | 0.01 | 0.02 | 84s |
**Result:** NO - P/L **-4.89** - bankroll after $971.97 ### 22:45-23:00 UTC - `KXBTC15M-26NOV051800` - settled **NO** Strike **75,723.21** - spot at open 75,685.96 - spot at close 75,580.01 - final reference 75,601.94 **Decision (auto_committee)** at 22:48:15 UTC: **NO** 6 contracts @ 0.77 ($4.62, fee $0.08), edge +9.5c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.25, conf 0.61): Spot down over the last minutes: 3m -0.191% (-1.99 sigma), 5m -0.251% (-2.02 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.59, conf 0.33): Spot is stretched below its 10-minute mean by -0.148% (-0.85 sigma); reversion favors YES. - **Quant desk** (Opening, P(yes)=0.11, conf 0.52): Brownian fair value: spot is -1.22 sigma from the 75723.21 strike with 11.9 min left and 0.215% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.56, conf 0.41): Book is bid-heavy (imbalance +0.40) and the 2-minute tape shows net taker selling of YES (-0.22); spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.25, conf 0.61): Flow is not confirming the move, but price is the primary signal. Standing pat. - **Mean-reversion desk** (Rebuttal, P(yes)=0.59, conf 0.20): Trend score -2.00 with flow confirming. Fading a live trend is a low-probability game - I stand down. - **Quant desk** (Rebuttal, P(yes)=0.11, conf 0.52): Model and market are broadly aligned; the edge is thin but real. - **Microstructure desk** (Rebuttal, P(yes)=0.56, conf 0.41): Flow is mixed; no strong conviction from my side. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.12, conf 0.38): Quant anchor P(yes)=0.11, tilted -0.008 by the discretionary desks, shrunk to 0.12 for model error; 72% weighted agreement, 0.38 confidence. I take NO at 0.77 x 6 contracts (4.62 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 22:45:29 | 75,685.96 | 0.41 | 0.41 | 871s | | 22:47:05 | 75,680.10 | 0.41 | 0.35 | 775s | | 22:48:41 | 75,549.55 | 0.17 | 0.15 | 679s | | 22:50:05 | 75,567.37 | 0.23 | 0.14 | 595s | | 22:50:53 | 75,504.05 | 0.10 | 0.06 | 547s | | 22:51:41 | 75,566.68 | 0.18 | 0.14 | 499s | | 22:52:35 | 75,595.13 | 0.18 | 0.19 | 444s | | 22:54:11 | 75,565.25 | 0.11 | 0.13 | 348s | | 22:55:47 | 75,667.61 | 0.33 | 0.35 | 252s | | 22:57:23 | 75,647.76 | 0.17 | 0.23 | 156s | | 22:58:59 | 75,540.82 | 0.01 | 0.00 | 60s |
**Result:** NO - P/L **+1.30** - bankroll after $971.97 ### 23:00-23:15 UTC - `KXBTC15M-26NOV051815` - settled **NO** Strike **75,601.94** - spot at open 75,537.77 - spot at close 75,284.76 - final reference 75,322.01 **No trade (fast committee).** First checkpoint at +68s: best net edge +0.001 is below the 0.04 threshold (P(yes)=0.32, yes 0.34/0.35). 12 evaluation(s) this window; the desk stayed flat. ### 23:15-23:30 UTC - `KXBTC15M-26NOV051830` - settled **YES** Strike **75,322.01** - spot at open 75,284.85 - spot at close 75,767.44 - final reference 75,782.09 **No trade (fast committee).** First checkpoint at +66s: best net edge -0.002 is below the 0.04 threshold (P(yes)=0.48, yes 0.50/0.51). 12 evaluation(s) this window; the desk stayed flat. ### 23:30-23:45 UTC - `KXBTC15M-26NOV051845` - settled **YES** Strike **75,782.09** - spot at open 75,764.74 - spot at close 75,951.23 - final reference 75,937.74 **No trade (fast committee).** First checkpoint at +68s: best net edge +0.037 is below the 0.04 threshold (P(yes)=0.40, yes 0.34/0.35). 12 evaluation(s) this window; the desk stayed flat. ### 23:45-00:00 UTC - `KXBTC15M-26NOV051900` - settled **YES** Strike **75,937.74** - spot at open 75,966.43 - spot at close 75,979.55 - final reference 75,975.89 **Decision (auto_committee)** at 23:47:16 UTC: **YES** 8 contracts @ 0.60 ($4.80, fee $0.14), edge +5.6c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.65, conf 0.46): Spot up over the last minutes: 3m +0.081% (+0.93 sigma), 5m +0.181% (+1.61 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.42, conf 0.31): Spot is stretched above its 10-minute mean by +0.116% (+0.73 sigma); reversion favors NO. - **Quant desk** (Opening, P(yes)=0.67, conf 0.44): Brownian fair value: spot is +0.44 sigma from the 75937.74 strike with 12.8 min left and 0.195% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.55, conf 0.42): Book is offer-heavy (imbalance -0.20) and the 2-minute tape shows net taker buying of YES (+0.46); spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.65, conf 0.48): Order flow points the same way as price. I hold my conviction. - **Mean-reversion desk** (Rebuttal, P(yes)=0.42, conf 0.18): Trend score +1.17 with flow confirming. Fading a live trend is a low-probability game - I stand down. - **Quant desk** (Rebuttal, P(yes)=0.67, conf 0.44): Model and market are broadly aligned; the edge is thin but real. - **Microstructure desk** (Rebuttal, P(yes)=0.55, conf 0.42): Flow is mixed; no strong conviction from my side. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.67, conf 0.23): Quant anchor P(yes)=0.67, tilted +0.013 by the discretionary desks, shrunk to 0.67 for model error; 93% weighted agreement, 0.23 confidence. I take YES at 0.60 x 8 contracts (4.80 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 23:45:34 | 75,966.43 | 0.55 | 0.61 | 866s | | 23:46:58 | 75,949.24 | 0.54 | 0.58 | 782s | | 23:48:22 | 75,961.62 | 0.53 | 0.64 | 698s | | 23:49:46 | 75,920.73 | 0.47 | 0.44 | 614s | | 23:51:10 | 75,889.61 | 0.41 | 0.30 | 530s | | 23:52:34 | 75,825.35 | 0.17 | 0.13 | 446s | | 23:53:58 | 75,765.29 | 0.09 | 0.05 | 362s | | 23:55:22 | 75,823.53 | 0.11 | 0.08 | 278s | | 23:56:46 | 75,831.17 | 0.11 | 0.13 | 194s | | 23:58:10 | 75,917.47 | 0.52 | 0.65 | 110s | | 23:59:34 | 75,973.83 | 0.99 | 0.86 | 26s |
**Result:** YES - P/L **+3.06** - bankroll after $971.97 ### 00:00-00:15 UTC - `KXBTC15M-26NOV051915` - settled **YES** Strike **75,975.89** - spot at open 75,941.12 - spot at close 76,012.46 - final reference 76,014.08 **No trade (fast committee).** First checkpoint at +74s: best net edge -0.016 is below the 0.04 threshold (P(yes)=0.55, yes 0.54/0.55). 12 evaluation(s) this window; the desk stayed flat. ### 00:15-00:30 UTC - `KXBTC15M-26NOV051930` - settled **YES** Strike **76,014.08** - spot at open 76,011.07 - spot at close 76,083.38 - final reference 76,085.23 **No trade (fast committee).** First checkpoint at +66s: best net edge -0.013 is below the 0.04 threshold (P(yes)=0.67, yes 0.66/0.67). 12 evaluation(s) this window; the desk stayed flat. ### 00:30-00:45 UTC - `KXBTC15M-26NOV051945` - settled **YES** Strike **76,085.23** - spot at open 76,107.79 - spot at close 76,272.32 - final reference 76,267.68 **Decision (auto_committee)** at 00:31:16 UTC: **YES** 10 contracts @ 0.47 ($4.70, fee $0.18), edge +4.9c.
Committee debate - **Momentum desk** (Opening, P(yes)=0.46, conf 0.30): Spot down over the last minutes: 3m -0.061% (-0.65 sigma), 5m +0.041% (+0.34 sigma). - **Mean-reversion desk** (Opening, P(yes)=0.50, conf 0.20): Spot is stretched above its 10-minute mean by +0.004% (+0.03 sigma); reversion favors NO. - **Quant desk** (Opening, P(yes)=0.53, conf 0.41): Brownian fair value: spot is +0.09 sigma from the 76085.23 strike with 13.8 min left and 0.209% 15m vol. - **Microstructure desk** (Opening, P(yes)=0.58, conf 0.32): Book is bid-heavy (imbalance +0.10) and the 2-minute tape shows net taker buying of YES (+0.21); spread 1.0c. - **Risk officer** (Opening, conf 0.95): No risk vetoes; trade is inside the risk budget. - **Momentum desk** (Rebuttal, P(yes)=0.46, conf 0.24): The quant model sits at P=0.53, against my trend read. Two different lenses, so I cut conviction rather than fight it. - **Mean-reversion desk** (Rebuttal, P(yes)=0.50, conf 0.20): Price is still stretched from its mean; I keep my fade, sized small. - **Quant desk** (Rebuttal, P(yes)=0.53, conf 0.35): Both discretionary desks lean against fair value. Either my vol estimate is stale or the market knows something - I widen my error bars. - **Microstructure desk** (Rebuttal, P(yes)=0.58, conf 0.32): Flow is mixed; no strong conviction from my side. - **Risk officer** (Rebuttal, conf 0.95): No objection. Keep stake inside the risk budget. - **Chair** (Ruling, P(yes)=0.54, conf 0.05): Quant anchor P(yes)=0.53, tilted +0.004 by the discretionary desks, shrunk to 0.54 for model error; 73% weighted agreement, 0.05 confidence. I take YES at 0.47 x 10 contracts (4.70 notional).
Extracted market path | time (UTC) | spot | YES mid | model P(yes) | left | |---|---|---|---|---| | 00:30:38 | 76,107.79 | 0.48 | 0.55 | 861s | | 00:32:02 | 76,098.21 | 0.51 | 0.52 | 777s | | 00:33:26 | 76,139.67 | 0.64 | 0.66 | 693s | | 00:34:50 | 76,214.38 | 0.80 | 0.85 | 609s | | 00:36:14 | 76,202.07 | 0.81 | 0.88 | 525s | | 00:37:38 | 76,198.23 | 0.83 | 0.85 | 441s | | 00:39:02 | 76,198.39 | 0.84 | 0.88 | 357s | | 00:40:26 | 76,210.66 | 0.90 | 0.92 | 273s | | 00:41:50 | 76,240.69 | 0.97 | 0.99 | 189s | | 00:43:14 | 76,218.17 | 0.99 | 1.00 | 105s | | 00:44:38 | 76,256.24 | 1.00 | 1.00 | 21s |
**Result:** YES - P/L **+5.12** - bankroll after $971.97 ### 00:45-01:00 UTC - `KXBTC15M-26NOV052000` Strike **76,267.68** - spot at open 76,261.87 **No trade (fast committee).** First checkpoint at +75s: best net edge -0.014 is below the 0.04 threshold (P(yes)=0.52, yes 0.51/0.52). 10 evaluation(s) this window; the desk stayed flat. ## Method - **Data**: Kalshi public REST (`/trade-api/v2`) for market, order book and prints; BTC spot is a clustered consensus of Coinbase / Kraken / Binance sampled every 5s, with single-tick glitches rejected and a 3-tick median used for fair value so one noisy quote cannot swing the model. - **Committee**: four directional desks (momentum, mean-reversion, quant, microstructure) each state P(settle YES); the risk officer can veto (spread, time left, price band, daily loss breaker, thin book, volatility regime break). Round 1 opening statements, round 2 rebuttals that adjust conviction, then the chair anchors on the quant fair value, adds a capped discretionary tilt, and shrinks the result 5% toward 0.5 for model error. - **Entry**: only when the fair value beats the *ask you can actually pay*, after the 7%-of-P(1-P) fee, by at least 4c; inside a 0.15-0.85 mid band; with >= 120s left; one position per window. The desk stands down when the quote is collapsing (>6c in 60s against the entry side), on a >3-sigma one-minute shock, or when the 'edge' exceeds 18c (that is a stale quote or a broken vol input, not free money). - **Exit**: hold to settlement. A 400-window candle backtest of the same entries showed the stop/flip/time-stop rule set cut net P/L from +$567 to +$60 - early exits were selling binaries at local extremes on noisy model flips. Risk is bounded by the 1% stake, so positions ride to expiry. Trades taken before this change keep their historical exit reason in the log for comparison. - **Fees**: `ceil(0.07 * contracts * price * (1 - price))` per side, as published by Kalshi.