How the strategy rating works

One number, four questions

Two strategies sit side by side on the leaderboard, both rated 72. One is a patient grinder with a modest edge and a flat, boring profit curve. The other is a violent ride that happens to be up this week. Same number, completely different animals. That is the first thing to understand about the rating: it runs from 0 to 100, it looks like a hotel review (80 good, 60 average, 40 avoid), and it must not be read like one.

The rating is a composite built from four sub-scores, each on its own 0 to 100 scale, and each answers a different question about the strategy. The composite tells you a strategy is worth opening. The sub-scores tell you what you are actually looking at. Treat the rating as a shortlist tool, then open the detail page and read the four components before you commit money or attention.

The four sub-scores

• Performance. Is the strategy making money? Everyone looks here first, and it is the least special score of the four. Profit is what you want, but profit over a short stretch is mostly noise.
• Reliability. How steady are the results, and over how many bets? A strategy that grinds out its yield across hundreds of picks scores higher than one that made the same profit in one lucky fortnight. Sample size lives here: a thin history caps this score no matter how pretty the curve looks.
• Drawdown. How deep does the pain get? This scores the worst peak-to-trough losses. Two strategies can end at the same profit while one of them spent months underwater on the way. If you would have quit during that stretch, the end result was never available to you.
• Freshness. Is the strategy still alive? This one is driven by the days since the last pick. Markets move, leagues change format, and a filter tree tuned to conditions that no longer exist simply stops firing. A strategy that has not produced a pick in a long time may be a museum piece, and freshness is the score that says so.

Why CLV sits next to the rating, not inside it

Closing line value is deliberately not one of the four sub-scores. Simulated backtest bets are priced at the closing odd by default, so CLV only exists on live picks. Folding it into the composite would either reward nothing on a young strategy or punish it for data that cannot exist yet. So every strategy shows its average CLV against Pinnacle's closing line as a stat of its own, and the leaderboard sorts on it. Read it alongside the rating, because it answers the one question the rating cannot: is this strategy skilled, or merely lucky so far?

Results lag skill, and they lag badly. Over thirty or fifty bets, a genuinely sharp strategy can lose money and a coin-flip strategy can look brilliant. That is variance doing what variance does. Closing line value cuts through the wait: every single bet either took a better price than where the market closed, or it did not. No luck enters the measurement, and you do not need a season's worth of results before the signal shows up.

Pinnacle's closing price is the sharpest public estimate of a match's true probability, so consistently beating it means the strategy finds prices the market later agrees were too generous. A strong average CLV with mediocre recent profit usually means the strategy is early: the results have not caught up with the skill yet. The reverse pattern, healthy profit with poor CLV, is the one to distrust, because that is exactly what running hot looks like. The full argument is in Value betting and CLV.

Two scopes: Since shared and Overall

Every stat on the platform exists in two versions. Overall includes the backtest, meaning the simulated bets generated when the creator tested the rules against history. Since shared counts live picks only: bets detected on real upcoming matches after the strategy went public.

The public leaderboard ranks on the Since shared scope by default, and that choice exists to protect you. A backtest can be tuned, knowingly or not, until it fits the past perfectly, which is exactly why spectacular backtests so often disappoint live. Live picks cannot be curve-fitted. The strategy called the future or it did not. Ranking on live results only means nobody climbs the leaderboard on the strength of a story about history.

The honest flip side: a freshly published strategy has almost no live bets, so its Since shared numbers, and its rating with them, swing hard with every result. That volatility is not a flaw in the strategy. It is a small sample behaving like a small sample. Give young strategies time before you conclude anything, in either direction.

Sorting past the rating

The rating is the default sort, not the only one. The leaderboard also sorts by yield, P&L, max drawdown, bets, followers, average P&L per month, CLV score and creation date, and combining them is where the real reading happens. A few patterns worth internalising:

• High yield on few bets means unproven. Sort by yield and the top is always crowded with tiny samples. Check the bets column before you get excited; 40% yield over 25 picks is an anecdote.
• High rating, low freshness: check the pulse. A strategy can coast on a strong record while having quietly stopped firing. Look at the days since the last pick before you commit attention to something that may already be retired.
• CLV score is your early-detection sort. It surfaces strategies whose skill is measurable even though their profit curve has not had time to prove it.
• Max drawdown is your pain filter. Sorting by it shows which strategies have historically demanded the strongest stomach. Match this to yourself honestly.
• Followers measure popularity, not edge. A crowd can be early or a crowd can be wrong; the number tells you neither.

One practical note: browsing all of this, the leaderboard and every public strategy's stats, is open to anyone. Actually following a strategy and receiving its picks requires an active plan; see the plans page.

What a rating cannot promise

A rating is a summary of the past, computed honestly. It is not a forecast. Edges decay, leagues change, and the market gets sharper every year, so a strategy's 85 tells you what it did, not what it will do. And because young strategies carry thin Since shared samples, their ratings are the most volatile numbers on the site: a newcomer's 78 and a veteran's 78 are not the same claim. Use the rating to decide what deserves a closer look. Use the sub-scores, the scopes and your own tolerance for drawdown to decide what deserves your follow.

Where to go next

For how to act on all this before clicking Follow, read Following strategies well. For why small samples lie so convincingly, read Variance and sample size. And for the deep case on CLV itself, read Value betting and CLV.