Each coin gets one locked forecast, then the live model checks whether that original direction holds. Compare early versus confirmed prices without changing the original call. No trades are placed.
ConnectingWaiting
NOT REAL-MONEY READY: probabilities are untrained and NOT calibrated; do not mistake a stable pick for a proven edge. Indicative spot differs from the CF Benchmarks index used to settle Kalshi contracts. PAPER PICK means a hypothetical selection, not a real-money order recommendation.
Checking USD price feeds…
Feed diagnostics (tap if any coin is stuck)
Loading feed status…
0/5
Active windows
0/5
Models ready
0
Window paper picks
—
Paper win rate
--:--
Next rollover
OUTCOME FORECASTS · LOCK STATUS (tap to collapse)
Locked outcome forecasts YES or NO per ready coin
0/5 locked forecastsNot automatic trades
Locks from minute 6 when a recorded minute-3 direction agrees, otherwise waits until minute 9 if fresh data is available. One YES/NO forecast per coin; it never flips. Weak forecasts are still weak, and a late research-only lock can recover through 1:45 remaining if the feed was briefly unavailable. No lock occurs on missing data or after that cutoff. These percentages are unvalidated estimates, not established win rates.
Checking opening data…
Entry Timing Lab Price versus confirmation — not a trading instruction
Is the original forecast holding up?Observed at 3 / 6 / 9 / 12 min
Each row compares the locked forecast with the current model outlook, original versus current contract ask, and fee-adjusted research margin. A locked YES/NO will never change, but the fresh outlook can contradict it. Missing checkpoints stay missing—no backfilled data.
Waiting for locked forecasts to evaluate shared market risk.
Loading entry timing observations…
Early versus late entry observations and settled outcomes
Collecting real-time price checkpoints before settlement.
Hypothetical return examples use observed asks and estimated fees, not actual fills. Different minutes can contain different samples; only paired results are directly comparable. No automatic learning or staking increases occur.
Important: A later confirmation is not a reason to double down. Five YES contracts may be five expressions of one correlated crypto-market view. Uncalibrated percentages cannot justify a large combination position.
Best setup right now1 primary · 2 alternatives
An easy-to-scan ranking of paper opportunities. A means locked paper candidate still passes current quote checks; B means watch; C means pass. None is a verified buy signal.
CHECKING MARKET DATA
Waiting for fresh quotes and charts.
Tier C — currently not qualifying (including pending)
Avoid treating five correlated YES bets as five independent bets. Never increase a position just because the same uncalibrated model stays bullish.
Mistake tracker & timing study Prospective, not backfilled
Do forecasts work better later?0 settled · 0 awaiting
Checking stored and pending checkpoints…
Records actual live predictions near minutes 3, 6, and 9 and later checks Kalshi’s official YES/NO result. Unsettled checkpoints now appear as saved and pending rather than being counted as zero observations. If the page was closed, a missed checkpoint is not reconstructed from later prices. The model is not automatically retrained on a few losses.
Collecting prospective samples before comparing timing choices.
Recent incorrect locked forecasts
No graded forecasts yet.
What failed: condition counts
Learning study not started.
Storage is local to this site and may be cleared in Private Browsing. Save a JSON backup before creating a new pages.dev project. On your old V12 site, use “Export locked forecasts CSV” and import that CSV here to preserve earlier picks. Imported outcomes are rechecked with Kalshi when available. History is checked for settlements only while this page is open.
Locked Top 3 paper picks (full audit details)
Top 3 paper opportunities 0–3 per 15-minute window
Locked selections0/3 locked
Ranked by estimated fee-adjusted margin, uncertainty and spread. Highly correlated same-direction positions cannot occupy multiple slots. An unvalidated model does not establish an actual edge.
Collecting live data…
When will a trade qualify? Separate from forecasts
Checking...Cutoff pending
A strong candidate requires sustained confirmation. No pick is guaranteed.
BTC
WAITING
Loading model
--:--
ETH
WAITING
Loading model
--:--
SOL
WAITING
Loading model
--:--
DOGE
WAITING
Loading model
--:--
XRP
WAITING
Loading model
--:--
All five markets Tap a coin for charts and reasoning
Individual marketsScreening
Last five 15-minute windows Actual target resets
Kalshi's target does not sway within a window. It resets at each rollover. These are the recent published targets and final YES/NO results when settled. Missing results are marked pending.
Checking previous windows…
Forward-test validation Requires settled outcomes
NOT VALIDATED FOR REAL MONEY
Every directional forecast after the opening observation period is time-stamped and locked locally. After settlement we compare its predicted YES percentage with the actual outcome. Brier score: lower is better. Kalshi's market midpoint is the benchmark, not a perfect truth source. Paper returns include estimated general taker fees but not slippage or guaranteed fills.
0Settled forecasts
0Distinct windows
—Model Brier (lower better)
—Market Brier
0Settled paper picks
—Hypothetical ROI after fees
Collecting forward-tested predictions.
Probability calibration by bin
No outcomes yet.
Paper results & data export
Waiting for paper signals. A pick is logged only after the sustained confirmation gate passes. Results are checked when the page is open. Does not represent actual fills.
0 logged
How the Top 3 system works
Trade-entry model: each coin gets at most one optional trade-value paper pick: EVALUATING, PAPER PICK YES, PAPER PICK NO, or SKIP WINDOW. Across all coins, no more than 3 picks can lock in one 15-minute window. Same-direction contracts with strong historical correlation are screened as overlapping exposure. A pick may be issued at any point after the first minute, but no later than 1:45 before expiration. It needs three independent readings spanning at least 55 seconds, so a new candidate must appear earlier than that cutoff. A prediction never changes sides in the same contract; deteriorating conditions trigger a warning, not another pick. MONITOR YES/NO: an informational underdog alert, not a new trade prediction. REVERSAL WATCH: momentum is improving toward a flip. REVERSAL FORMING: momentum and Kalshi quote movement agree. PAPER REVERSAL: a stricter, unvalidated value screen for backtesting only. An underdog is not automatically a bargain. Market outlook: LEAN YES, LEAN NO, or NEUTRAL from the chart model—even when trading is unattractive. Trade-value screen: PAPER TEST YES/NO, WATCH, WAIT FOR PRICE, or NO ENTRY after estimated fees and uncertainty. The watch price is an experimental threshold, not a fair-value guarantee. A directional outcome forecast is separate from a paid trade. The model probability shown is raw and uncalibrated, NOT a verified chance of winning. An upper confidence bound is not claimed: forecast ranges are just sensitivity scenarios.
