How we measure, and what the measuring found
Any tool can show you a win rate. This page is the method behind ours, and the results it produced — which in several cases were bad. Those are on this page for the same reason the good ones are.
Why a win rate is not evidence
Suppose a rule wins 60% of the time with a 1.5-to-1 reward. That sounds like an edge. But buying at the open of a random session, with the same stop, the same target, the same time limit and the same charges, also wins a good share of the time — because the market drifts and the brackets are what produce the win rate, not the rule.
So the only number worth reporting is the difference between the rule and that random baseline, with an honest standard error on it. That is what we compute, and it is what decides whether a rule is allowed to be acted on at all.
The method
- Replay the rule. Every signal it would have produced, on real stored bars, with its own stop in ATR, target in multiples of risk, and time stop.
- Resolve bars pessimistically. If a bar touches both the stop and the target, the stop wins. An entry fills at the next bar’s open, never at the signal’s own close.
- Apply the full Indian cost model. Brokerage, STT at the right rate for the instrument and product, exchange transaction charges, SEBI fees, stamp duty and GST.
- Replay the baseline. The identical bracket, bought at the open of random sessions near each real trade.
- Report the difference and its standard error, walk-forward: fit on one window, test on the next, roll.
The four labels
Every rule in the product carries exactly one of these, computed the same way.
| Label | Means | Can it be acted on? |
|---|---|---|
| favourable | Beat the baseline by at least two standard errors, over twenty or more finished trades. | Yes — this is the only label that qualifies. |
| unproven | Ahead of the baseline, but inside the noise. | No. Shown with its figures so you can judge it yourself. |
| unfavourable | Lost money, or did no better than random entries. | No — and it stays visible, because a failure is a result. |
| unmeasured | Fewer than eight finished trades, or never replayed at all. | No. Labelled everywhere it appears, because this is the state people assume has been tested. |
What we found
Replayed over a hundred-odd NSE stocks, five years of daily bars, walk-forward, net of costs. These are our own results and they are not flattering.
| Rule family | Result after costs | Versus random entries | Verdict |
|---|---|---|---|
| Opening-range breakout | −0.68% per trade | 0.96% below random (±0.10) | unfavourable |
| Composite buy call | −0.14% per trade over 3,738 trades | −0.03% (±0.14) — indistinguishable | unfavourable |
| Price-action triggers | −0.05R | −0.12R below random (±0.02) | unfavourable |
| Four charts agreeing (as an intraday filter) | Reduced returns rather than improving them | Negative contribution | unfavourable |
Yes, we are telling you our rules did not beat random
We could have shown you the win rates instead. 20% of those buy calls reached their target and the hit rates looked reasonable in sample — it is only the comparison against random entries, after costs, walk-forward, that shows there was no edge.
That is exactly why the comparison exists, and why the product refuses to act on these rules. A tool that hid this would be more appealing and less useful. The reads are on screen as reads; the product does not tell you to take them.
What this means for you
Use Stockhub to look at a market carefully — the four-timeframe reads, the chains, the structure, the screener — and treat every rule’s record as the reason to be sceptical of it. Where the record is bad, that is information about the rule, and it is more useful than a confident number would have been.
Results also go stale. A rule measured on five years of history describes those five years. We re-run the replays, and a rule whose rolling expectancy falls below a threshold switches itself off.
What we do not claim
- No returns are promised. Not by the product, not on this site, not in any conversation with us.
- No forecast is a prediction. A target ladder is a set of levels with historical reach times attached, not a statement about what will happen.
- Simulated is not real. Paper trading applies real costs but cannot model the slippage, partial fills and liquidity you would actually meet.
- Option strategies are the least evidenced part. Historical data for expired option contracts is not available from our data provider, so option-leg results are modelled from a pricing formula rather than measured on real chains. They are marked as such wherever they appear.