1. On the daily chart, draw three exponential moving averages: EMA 5, EMA 8, and EMA 13.
2. Wait until the close prints under all three EMAs for at least 3 consecutive bars. This is the "compression" — price is stretched below its short-term moving averages.
3. Take the trade on the bar where the close jumps back above all three EMAs in a single candle — i.e., the close is above max(EMA5, EMA8, EMA13).
4. Enter at 17:30 TR — 30 minutes before the BIST close. (We use 17:30 instead of 17:55 so the operator has time to confirm the setup before the bell.)
5. Manage next session: with high probability the morning open prints +0.45–0.50% above entry. Take profit there or trail the stop. Hard stop is −1.0% below entry.

We report both. They answer different questions.

• Touch rate — of all next-day bars after a signal, what fraction had a high reaching +0.45% at any point? This is what an operator watching the morning session can capture by exiting the moment the level prints. It's the "85–90%" claim.
• Strict win rate — of resolved signals, what fraction reached +0.45% TP before −1% SL on the next bar? Daily OHLC hides intraday path. When both are reachable inside a single bar, we conservatively assume SL fired first. This is the worst-case backtest for an inattentive trader who lets the stop do its job.

The truth lives in the gap between these two. Be honest about which one applies to you.

• Source: isyatirim.com.tr public HisseTekil endpoint (split/dividend-adjusted daily OHLC). No auth, no key.
• Latency: end-of-day, ~15 minutes after BIST close. Acceptable for a daily-timeframe strategy.
• Universe: BIST 30 + a curated set of liquid BIST 100 names. Thinly-traded stocks gap past TP/SL on volume spikes.
• We do not get an explicit "open" price (the endpoint exposes close, high, low, and adjusted weighted-average price). The TP/SL evaluation only needs high and low — both are real. The "open" stand-in is purely cosmetic for the gap-up badge.

Next.js 15 app. SQLite-backed daily candle store + signal ledger. The strategy engine is lib/strategy.ts — pure functions over candle arrays. Refresh pipeline pulls universe candles, runs the backtest from full stored history, and upserts signal outcomes idempotently. Charts are lightweight-charts (TradingView's open library).

• Refresh fires on demand via POST /api/refresh (or schedule a curl from cron).
• Want to track different tickers? Edit lib/universe.ts and refresh.
• Want to tune TP/SL or streak length? Edit DEFAULT_PARAMS in lib/types.ts, refresh, and watch the win rates move.

Educational. Not financial advice. Backtests describe the past, not the future. The strategy was supplied as a rulebook — we do not endorse it, we just measure it honestly.