STRATEGIES

Top‑5 picks, exit‑rule siblings & live scorecards

Back to Dashboard

Top 5 High-Probability Picks

Ranked by composite ML next-day return probability (XGBoost + LightGBM ensemble).

No prediction data available yet.

Picks Scorecard

How yesterday's 25 picks worked out — entry price vs today's live O/H/L & current price.

No archived picks yet. Picks are snapshotted each trading morning and scored after 3:30 PM close.

Strategy: Top‑5 by P(+3%) × volatility

Today · 06 Aug 2026

Our strategy: each morning buy the 5 stocks ranked by a 50/50 blend of next‑day P(+3%) and volatility (ATR + range), then hold to the close.

No prediction data available yet.

Strategy Scorecard

How the Top‑5 by P(+3%) × volatility actually played out — entry = pick‑day close, result = next session's close.

Not enough prediction history yet — the scorecard fills in as daily picks accumulate.

Strategy 3% · 3:20 PM

intraday
Today · 06 Aug 2026

At 3:20 PM the live price is taken as today's close and the model re‑ranks the Top‑5 by P(+3%) × volatility — so you can buy before the 3:30 close, not the next day.

No 3:20 PM snapshot yet — it's captured automatically each trading day at 3:20 PM IST.

Strategy 3% · 3:20 PM Scorecard

How the 3:20 PM Top‑5 played out — entry = 3:20 PM price, graded by how far the next day's HIGH ran above entry (E→H). Built forward day‑by‑day (no historical intraday data to back‑test).

No 3:20 PM picks graded yet — results appear the morning after the first snapshot.

Strategy: Top‑5 + −6% Disaster Stop

Today · 06 Aug 2026

Same Top‑5 picks as the hold‑to‑close card, but each position is exited if it craters past −6% intraday (gap‑aware) instead of held blindly to the close — pure tail insurance.

 Win rateAvg / pick≥ +2%
Hold‑to‑close
+ −6% Stop

No prediction data available.

Exit‑timing walk‑forward (30‑day & 60‑day OOS): the loose −6% stop was the only exit that beat plain hold‑to‑close on both windows while keeping win‑rate and the +2% hit‑rate identical — it just trims the rare crater. Tighter −3%/−4% stops tested worse.

Strategy: Top‑5 + 2% Trailing Stop

Today · 06 Aug 2026

Same Top‑5 picks as the hold‑to‑close card, but once a position first runs +2% it locks in by trailing a 2% stop below its running peak — protecting gains while still letting winners run.

 Win rateAvg / pick≥ +2%
Hold‑to‑close
+ 2% Trail

30‑day OOS backtest on real 5‑minute paths · trailing lifts win‑rate

No prediction data available.

Of the exit rules tested on real 5‑minute paths (30‑day & 50‑day OOS), arming a 2% trail after +2% was the only trailing variant to beat hold‑to‑close on both windows — its edge is a far higher win‑rate (more green days), trading a little of the fat‑tail upside for consistency. Requires intraday execution; statuses above are live approximations from the running session high.

Strategy: Top‑5 + 3:20 PM Exit

Today · 06 Aug 2026

Same Top‑5 picks, but instead of holding into the 3:30 closing auction you sell at ~3:20 PM — sidestepping last‑10‑minute volatility and settling before the auction.

 Win rateAvg / pick≥ +2%
Hold‑to‑close
3:20 PM Exit

30‑day OOS backtest on real 5‑minute paths · exit price = last 5‑min bar before 3:20 PM

No prediction data available.

Honest result: on these picks the closing 10 minutes are, on aggregate, slightly net‑positive, so exiting at 3:20 PM is a touch behind hold‑to‑close (). It is offered for those who prefer to be flat before the auction rather than as an edge.

Strategy Comparison · Top‑5 Exit Rules

30‑day OOS

The identical Top‑5 selection, graded on real 5‑minute paths under four exit rules — so you can see exactly what each rule trades off.

StrategyWinAvg/pick≥ +2%Cum
Hold‑to‑closebaseline
+ 2% Trailingbest win‑rate
− 6% Disaster stopbest cum
3:20 PM Exitflat by auction

All four hold the same picks — only the exit differs. 2% trailing wins on consistency (highest win‑rate); the −6% stop wins on total return by cutting only the rare disasters; 3:20 exit ≈ hold‑to‑close. Cum = compounded daily mean across ~ sessions; backtest uses xgb for fast relative comparison.