How InDecision Caught BTC's 4.7% Crash: A Live Bot Case Study

Theory is clean. Live trading is brutal. This is a live case study from February 23, 2026 — real positions, real P&L, real consequences.

By the time BTC closed at $64,455 that afternoon, InDecision had already told the story twice: once to show what a precision call looks like, and once earlier in the day to show what a trap looks like. Both lessons were visible in real time. Here's the breakdown.

The Macro Setup

The 4-hour chart coming into February 23 was not ambiguous. BTC had dropped from approximately $69,000 to $64,500 over the prior 48 hours — a −6.5% structural move with sustained selling volume. The 4h MACD read −325.54 on the MACD line and −589.94 on the signal, with a histogram of −264.40. This is not a correction. This is distribution. Price was tracking below all moving averages, clinging to the lower Bollinger Band, with no base forming.

The 15m timeframe at the open of trading that day confirmed the picture: MACD at −148.42 / −311.96 with a histogram expanding further negative. No bullish divergence. No volume capitulation. Just sustained one-directional pressure.

This is the environment InDecision was operating in. The macro context was unambiguous before a single trade was placed.

Trade 1: 10:08 AM ET — The Aligned Call

The first live trade of the day: BTC DOWN at approximately $0.49 (49¢) implied odds on a 15-minute binary window.

What made this trade correct wasn't just that it was DOWN. It's what the model saw:

• Price momentum: −0.41% intraday delta with directional confirmation across lookback windows
• Technical analysis: 15m MACD negative and extending, no counter-signal
• Volume confirmation: 2.0× normal 15m volume — not a low-conviction drift, active selling
• InDecision bias: NEUTRAL aligned — no contradiction from the daily feed
• Trade price: 49¢ for DOWN tokens = near-even odds on a trade with directional edge

The market pricing DOWN at 49% when it was actually trending down at the macro level is the edge InDecision is designed to find. You're not paying a premium for a late signal. You're getting fair odds on an aligned thesis.

Result: WIN. Net P&L: +$4.38.

The Bounce Trap: 11:21 AM ET

One hour and thirteen minutes later, the same bot placed a second BTC trade. This one went differently.

BTC UP at approximately $0.94 (94¢) implied odds. A brief 15-minute positive candle had fired in the middle of a confirmed macro downtrend. The 15m signal scored enough points to pass the threshold. The market priced UP tokens at 94 cents, implying 94% probability of an UP close. The bot bought in.

Let's be precise about why this trade was wrong before we know the outcome.

A 15m binary priced at $0.94 means you're risking $0.94 to win $0.06. To break even at fair value, the actual probability of UP needs to exceed 94%. In the middle of a day where BTC had already dropped 4%+, with the 4h MACD at −325 and continuing to trend down, what is the actual 15m UP probability? It is not 94%. It's somewhere around 50% — the noise floor of short-term intraday candles. The market was mispriced by momentum chasers, not InDecision.

The candle was a dead-cat bounce. A brief recovery in the middle of a crash is not reversal confirmation — it's a liquidity grab before continuation.

Result: Direction was technically correct (market closed UP). But at 10% fee rate, the gross profit of $0.06 per token is consumed entirely by fees. Net P&L: −$0.22. BREAK EVEN.

This trade revealed a gap in InDecision's extreme-odds handling that the team immediately patched.

The Rule That Changed

After this trade, we implemented a hard rule into the scoring system:

If the implied probability of the bet direction is ≥ 85%, skip the trade regardless of score.

The logic: a market priced at 90%+ for one outcome is not confirmation. It is late-momentum. Whoever is on the other side of a 94¢ bet isn't stupid — they know something. A healthy signal at fair odds is worth acting on. A crowded signal at extreme odds is worth skipping.

The updated formula applies a hard skip before any other evaluation: if bet_price >= 0.85: skip_reason = "Extreme market odds — bounce trap risk". No score override. No threshold creep. Hard stop.

Trade 3: ~12:15 PM ET — The Precision Call

With the extreme-odds rule live, the bot entered the afternoon session. BTC had continued its decline. The 15m MACD was now −149.73 / −313.80. The histogram had expanded to −163.87. Price sat at $64,529 — down 4.57% on the day.

The bot placed another BTC DOWN position. This time at fair odds. The DOWN token was trading at approximately 55¢ — 55% implied probability for the direction aligned with every macro signal on every timeframe.

Signal read at time of execution:

• Score: 89/100 (STRONG conviction threshold in conservative mode: 90, cleared in moderate/aggressive)
• Momentum: 30/35 — strong directional delta
• Technical analysis: 14/20 — MACD confirmation across timeframes
• Volume: 2.0× normal — active selling pressure confirmed
• Trade price: $0.555 per DOWN token (fair odds)

This is the trade InDecision is built for: aligned bias, fair odds, multi-timeframe confirmation, high conviction score.

At the time of writing this case study, the position is still open. BTC has continued lower. The trade is in profit.

What This Proves

Three trades, three different outcomes, one consistent framework.

The $4.38 WIN and the $0.22 BREAK EVEN were placed in the same market on the same day on the same asset. The difference was not analysis quality — the bot's signal engine scored both trades comparably. The difference was odds quality. One trade had edge. One trade paid a premium to be late to a crowd signal.

InDecision is not a directional predictor. It's a signal quality filter. It tells you whether the signal you're seeing has actual information content or is just the market's emotional reaction to its own momentum. A 94% priced market is telling you the crowd has already acted. You're not getting the trade — you're getting the bag.

The Multi-Timeframe Rule

One lesson from this day that applies beyond InDecision:

Before placing any trade in a given direction, check the 1h and 4h MACD. If both are negative and expanding, reduce your UP conviction by 30-40 points mentally. Not because the 15m can't close UP — it can. But because you're fighting two higher timeframes with one candle. The expected value of counter-trend bets in trending markets is structurally negative.

InDecision's DualCaseAggregator was specifically designed to catch this. When the intraday bias feed (IntraCaseAggregator, 5-minute refresh) scores a NEUTRAL_BEARISH from RSI + MACD + Bollinger Band signals across multiple timeframes, that score gets injected into every 15m evaluation. The InDecision contribution to the total score penalizes counter-trend UP bets and amplifies aligned DOWN bets.

The bounce trap survived that filter because the InDecision score was positive before the extreme-odds rule was in place. Now it doesn't. The system learned. The rule is live.

Final Numbers (as of 12:35 PM ET)

Trade	Time	Direction	Odds	Score	Result	Net P&L
BTC DOWN	10:08 AM	↓	49¢	—	WIN	+$4.38
BTC UP	11:21 AM	↑	94¢	89/100	BREAK EVEN	−$0.22
BTC DOWN	~12:15 PM	↓	55¢	89/100	OPEN	pending

Total decided P&L: +$4.16. Total wallet balance: $47.98. Win rate on resolved positions: 100% (BREAK EVEN counted as win in direction; negative in P&L — which is why we track both).

The framework is live. The trades are real. The edge is measurable.