Best Entry Signal is now a good time to bet-and-hold to close?
--
--
Favored Side
--
Model Win Probability
--
Market Price (cost to enter)
--
Edge
--
Edge = model's estimated win probability minus the market's current price for that side. Positive edge means
the model thinks you're getting a discount; negative means the market has already priced this in (or more).
Expected value per $1 staked: --.
Unvalidated -- real evidence comes from the Automated Bets panel's accumulated results at different entry times, not this formula alone.
Leading signal: the market's displayed odds are a real order-book quote that can stay frozen for several
seconds while the underlying price keeps moving -- so a meaningful price move during a "stale" stretch may
show up here before the Up/Down % itself visibly updates. Unvalidated thresholds, first guess only.
Trend Regime is BTC skewing one direction more than normal 50/50 noise?
--
Lookback
Up/Down
Dominant
vs 50% coin flip
Based on actual settled window outcomes, not news classification -- if something (news or otherwise) is
big enough to skew many windows in a row, it shows up here directly. Small sample caveats apply heavily
early on; a real regime should hold up, not just appear for a couple of windows.
Price to Beat (target)
--
Window closes in
--:--
Current
----
Up · --%
Down · --%
--x
--x
Current Kalshi Window
Strike / Target
--
Yes Bid/Ask
--
No Bid/Ask
--
Status
--
CF Benchmarks BRTI (Kalshi settlement index)
--
Requires a Kalshi API key. Not yet configured -- this will populate once credentials are added.
Pyth Network BTC/USD independent raw feed, NOT the settlement source
--
Best Bid
--
Best Ask
--
vs BRTI
--
Added 2026-08-24 to test whether an independent, non-averaged price feed works better than BRTI for the
momentum/window-open-trend experiments -- BRTI showed no real edge there. Data is accumulating; see
backtest_time_to_win_pyth.py and analyze_window_open_trend_pyth.py once there's enough history.
Prediction (baseline heuristic, not a trained model yet)
--
Confidence
--
--
Swing Alerts
Session Replay full tick-by-tick history of a settled window, with notable moves highlighted
Yellow/green markers highlight the zigzag-detected notable swings (>=8c moves) plus the session's overall
high/low -- long unmarked stretches ARE the chop/consolidation periods, that's deliberate, not missing data.
Built 2026-08-24 after manually watching a live window tick-by-tick end to end.
Underdog Long-Shot Tracker does a $1-4 bet on the loser ever pay off?
Fires when the cheaper side (Up or Down, whichever has the lower ask) drops to 1-3 cents with 5 minutes
or less left in the window -- one event per window, at the first moment this happens. The ~2-minute mark
specifically is broken out below since that's the point being watched most closely, but the full curve
down to 5 minutes is tracked too so it's clear whether entering earlier/later actually matters.
Bucket
n
Win Rate (95% CI)
Avg Ask
Breakeven Win Rate
Edge
Breakeven win rate = the average entry ask price itself (a win at ask price p pays roughly 1/p per $1
staked, so you need to win more often than p% of the time just to break even). Edge = actual win rate
minus breakeven -- positive means this has actually been profitable at this sample size, negative means
it's a net loser even accounting for the big payout on a win.
Whale Alerts real bets, triggered by >= $25K taker trades
Fires when a single executed taker trade on the BTC 15m market is >= $25K.
Historically these large taker orders are on the winning side far more often than small ones.
The instant one fires, two real paper bets ($500 and $1,000) are automatically placed on the
whale's side and held to close -- at most one pair per window.
Bet Size
Wins/Total
Win % (95% CI)
Total P&L
After 1 Full Loss
vs 50% coin flip
Polymarket Whale Hedge Tracker paper only -- opposite side, equal contracts
Window
Kalshi
Polymarket
Contracts
Poly Entry
Results K/P
Combined P&L
Every qualifying Kalshi BTC whale trigger simulates buying the opposite outcome on Polymarket with the same
contract count, walking the live Polymarket ask book and applying its crypto taker fee. Kalshi settles on BRTI;
Polymarket settles on Chainlink's 60-second TWAP, so outcome disagreement is tracked explicitly.
Cross-Platform BTC 15m Scanner paper only -- KXBTC15M vs Polymarket, live ask-depth
Window
Pair
Matched
K VWAP
P VWAP
All-in
Locked P&L
Result K/P
Actual P&L
Walks Kalshi executable ask-side depth (reconstructed from the live order-book bid state) and Polymarket CLOB
asks for the same 15-minute window, evaluating Kalshi Up + Polymarket Down and Kalshi Down + Polymarket Up.
Records only pairs where same-result payout beats all-in cost by at least $0.002 per matched contract after fees.
Basis risk is real: Kalshi settles on BRTI while Polymarket settles on Chainlink's 60-second TWAP.
Captures the model's call, market-implied favorite, and executable asks at fixed time marks, then scores both
a $100 hold-to-settlement P&L after Kalshi's real taker fee. Used to compare the baseline heuristic against
the market's own favorite at each horizon.
SOL Late-Entry Experiment paper -- T-2:15, 0.15% target buffer
Window
Side
Entry
Outcome
P&L
Grew out of a cross-coin flip-rate study (2026-08-25): across BTC/ETH/SOL/HYPE/BNB/XRP/DOGE/ZEC/NEAR,
SOL showed the lowest rate of the T-2min leader losing (4.2% vs BTC's 10.8%), and after checking for
outlier-driven false positives (the same check that already caught two false BTC "edges" this project),
the only one with a genuinely positive median trade. No real whale-sized volume exists on SOL, so this
strategy doesn't copy anyone -- it just buys the leader directly, at $100/window, paper only until this
has a real track record. Filtered rows were recorded after the target-distance gate was added;
legacy rows predate it and are shown only for continuity.
SOL Late-Entry T-2:30 parallel paper model -- 0.10% target buffer
Window
Side
Entry
Target Distance
Outcome
P&L
Runs independently 15 seconds earlier than the existing SOL T-2:15 model with the same $100 stake,
0.10% SOLUSD_RTI target-distance gate, quote logic, fees, and hold-to-settlement scoring.
SOL T-2:15 vs T-2:30 Paired Comparison windows where both models traded
XRP Late-Entry Experiment paper -- T-2:15, 0.15% target buffer
Window
Side
Entry
Outcome
P&L
Second coin tried after re-running the same cross-coin methodology that validated SOL (2026-09-02):
real Kalshi candlesticks, real fees, and the same outlier/median check that caught illusory "edges" on
ETH/HYPE/ZEC. XRP passed every test SOL passed -- 3.5% T-2min flip rate (even lower than SOL's 4.2%),
positive median trade, and still solidly profitable after removing its 3 best trades. Same $100/window,
paper only until this has its own real track record.
XRP Late-Entry T-2:30 parallel paper model
Window
Side
Entry
Target Distance
Outcome
P&L
Runs independently 15 seconds earlier than the T-2:15 model, same $100 stake and 0.15% XRPUSD_RTI
target-distance gate -- included from day one since SOL's own data already showed T-2:30 nets more
total P&L than T-2:15 (cheaper entry price, negligible win-rate cost).
XRP Entry-Confidence Risk Check live paper data, T-2:15 + T-2:30 pooled
Target Distance
n
Wins
Win Rate
Total P&L
A one-off historical backtest (2026-09-02, 473 windows) found a much steeper win-rate gradient by entry
confidence for XRP than SOL ever showed -- as low as ~68% win rate in the least-decisive bucket, vs.
SOL's ~97% even in its worst bucket. This tracks the same risk check against XRP's own live paper
trades as they accumulate, to see if that pattern holds up in real (not backtested) data.
Live SOL Late-Entry Trading REAL MONEY -- T-2:15, 0.15% target buffer
--
Live Mode
--
Kill Switch
--
Cumulative Realized PnL
--
Stake / Trade
--
Window
Side
Contracts
Entry
Distance
Outcome
P&L
Places a REAL order on Kalshi whenever the SOL T-2:15 late-entry strategy fires -- entirely separate from the
paper $100 SOL late-entry trades above, which keep running regardless. Kill switch halts at -$500 cumulative
realized loss and does NOT auto-reset. Buys the currently-leading side only when SOLUSD_RTI is at least 0.15%
from the strike (raised from 0.10% after loss analysis: both filtered-era losses occurred below 0.15%, and
trades at >=0.15% distance were 163-0 in paper).
BTC Late-Entry Experiment paper only -- T-2:15, 0.25% target buffer
Window
Side
Entry
Target Distance
Outcome
P&L
Evaluates each BTC 15-minute window once at T-2:15. It enters the current leader only when BRTI is at least
0.25% from the target, then holds the $100 paper position to settlement. The historical test was 86-0 but
economically fragile, so this panel is collecting a forward-only track record before any live consideration.
BTC-FastScalp arb-bot momentum-lag scalper
Proxied from kalshi-arb-bot (EC2). BTC-FastScalp v2: window-open
continuation of the previous window's settlement side, 0.40-0.60 entry,
$5/15pp first-trade target, $1 second-trade, NO stops. Numbers shown are
TODAY ONLY by default; lifetime totals appear in muted text.
PAPER
Today PnL
--
Today Trades
--
Ticker
Side
Buy
Exit
Reason
P&L
LIVE
Today PnL
--
Today Trades
--
Status
--
Ticker
Side
Buy
Exit
Reason
P&L
Live Whale Trading REAL MONEY -- default OFF
--
Live Mode
--
Kill Switch
--
Cumulative Realized PnL
--
Stake / Trade
--
Window
Side
Contracts
Entry
Outcome
P&L
Places a REAL order on Kalshi (same account already used by kalshi-arb-bot's live trading) the instant a
qualifying whale trade fires -- entirely separate from the paper whale_500/whale_1000 strategies above,
which keep running regardless. Kill switch halts at -$20 cumulative realized loss and does NOT auto-reset.
Coordinate with kalshi-arb-bot before enabling if BTC-Trailing/BTC-FastScalp are ever live at the same time
(shared account/series -- see AGENTS.md).
Exit Probability Calculator rough estimate, not validated
Target Profit
Required Price
Chance of Reaching Before Close
"Chance of reaching before close" estimates the odds price moves far enough to sell for that profit
at some point before the window ends (not just at settlement) -- based on recent volatility and time
left, assuming no directional drift. It's a rough model, not validated against real outcomes yet.
Paper Trading (no real money)
Entered At (ET)
Window
Side
Amount
Entry Odds
Outcome
P&L
Automated $100 Paper Bets by Entry Time every minute, no user input -- empirical answer to "best entry time"
Every window automatically places a $100 paper bet at each elapsed minute from 1 through 14, on whatever the model's
prediction is at that moment, then holds to close. This is the actual empirical test of the entry-timing
question -- more reliable than the theoretical Edge/EV formula above once enough windows have settled.
Entry Time
Wins/Total
Win % (95% CI)
Total P&L
vs 50% coin flip
"Significant" means the result is unlikely to be pure noise at the current sample size -- it does NOT mean
the edge will persist or survive real transaction costs. Treat with real caution until it holds up across
many more windows.
Buckets trigger/resolved records by how far the reference index was from the strike and how many seconds
remained at trigger. Three fill models are shown: optimistic (triggering print capped at 100 contracts),
post-trigger (subsequent <=1¢ volume before close capped at 100), and queue-adjusted (post-trigger volume
minus a 100-contract queue hurdle, capped at 100). Legacy rows are excluded because their pre-close volume was
not captured cleanly.
Recent Settled Windows
Ticker
Strike
Result
Closed
BCH 15-Minute Predictor KXBCH15M -- same baseline heuristic + auto paper trades as BTC