ROI, strike rate and drawdown: the three numbers that tell you if a system is real

Strike rate feels honest, because it answers the question you actually asked: how often does this win? ROI feels like the real answer, because it tells you whether it made money. Neither one tells you whether a system is worth running. The number that actually decides that — drawdown — is usually the one nobody checks until they're already living through it.

The three numbers, plainly

• Strike rate — the percentage of qualifying bets that won. Says nothing about the price you were paid for winning.
• ROI — profit divided by total staked, settled at BSP net of commission. Says whether the method made money, but not what it cost you to find out, or what it will cost you to keep believing in it.
• Drawdown — the longest losing run, in points and in time. Says whether a real person, staking real money, would have made it to the other side.

Each one is a fair question. The mistake is answering only one of them and calling it a system.

Why strike rate lies on its own

A system that backs 1.30 favourites will win roughly three bets in four. A 75% strike rate feels like an edge. But at 1.30, break-even requires winning 76.9% of the time before commission — so a system that "only" hits 75% is losing money at a high, reassuring strike rate. Nobody staking it feels like they're losing, because they win most weekends. The P&L says otherwise.

Run the comparison the other way and the same trap flips: a system laying 8.0-shots might win only 82% of its lays and still print a healthy ROI, because losing 18% of the time at 8.0 costs far less than the 82% brings in. Strike rate on its own can't tell these two systems apart. It doesn't know what price it was collected at, so it can't tell you whether that price was worth taking.

Why ROI lies on its own

ROI fixes the price problem, but only once the sample is big enough for a single result to stop deciding the headline number. A system with 40 qualifying bets and one 50.0 winner in there can show +60% ROI that has nothing to do with whether the method works — it's one price move away from a loss. As a rough floor, be sceptical of anything under a few thousand qualifying bets, and always check what the number does with the best result removed. If pulling out the single best-priced winner turns a profitable system into a losing one, you haven't found an edge. You've found a lottery ticket that paid out once.

This is the same discipline covered in how to build a system that isn't curve-fitting: a positive number over a small, cherry-picked sample tells you almost nothing on its own.

Why drawdown is the one that actually ends staking plans

Take a system doing a genuine 3% ROI over 3,000 bets a year — a solid, defensible edge on paper. Buried somewhere in those 3,000 bets is a stretch of 60 or 70 points where nothing lands. If you don't know that stretch exists before you start, you find out live, on bet 40 of it, wondering whether the method just stopped working. Most people stop staking a genuine edge in exactly that window — right before the losing run ends, because they had no way to know it was a losing run and not a broken system.

Before trusting any rule set, look specifically at:

• Longest losing sequence — in points and in real time. Could you actually sit through it, with real money on?
• Month-by-month shape — is the profit a grind across the year, or three good months quietly carrying nine flat or losing ones?
• Recovery time — how long, historically, between a new equity low and getting back above the old high?

A system you abandon mid-drawdown because you didn't expect it isn't a staking failure. It's a research failure that happened three months late.

Reading all three together

A system only earns trust when all three line up: a strike rate that makes sense for the prices being taken, an ROI that survives having its best result removed, and a drawdown you've actually looked at in advance and decided you can live with. Two out of three is how promising systems get abandoned, and how bad ones get funded.