MabeKit Score
How much should you believe a filter the Strategy Cruncher found?
Every filter on a Strategy Cruncher report has a MabeKit Score from 0 to 100. It's the yellow MK badge on each card, and the MK Score button sorts the cards by it.
The score answers one question: how likely is this filter to keep working?
It's my take on how I choose between rules when I look at a report myself, boiled down to a single number so you can see at a glance which filters deserve a closer look.
What it measures
When I size up a rule, the profit it shows is the last thing I trust. The Cruncher found every filter by searching your backtest for the best split, so every filter looks good on that backtest. That's what it was chosen for.
What I want to know is whether the result looks like something real or like an accident of the data. Over many years of building strategies with these tools, I've developed a feel for which filters keep working and which fall apart. The score draws on many different factors to make that same call for every filter on your report.
A filter that looks like it will hold up scores high. A filter that looks great on profit but shows the warning signs I've learned to distrust scores low.
What it doesn't measure
The score deliberately leaves out how much a filter is worth. Profit factor, win rate and total profit, shown on each card, tell you that. The score tells you how much to believe those numbers.
Read the two together. A high score on a filter that barely improves anything is still a filter that barely improves anything.
A score of 0
A score of 0 means the filter keeps too few trades to judge. It doesn't mean the filter ranked last. Treat it as out, however good its other numbers look.
Reports from before mid-September 2026 also scored 0 for any filter whose kept trades lost money. If an older report on a losing backtest shows nothing but zeros, that's why. Crunch it again to get scores.
Using the score
There's no magic cutoff. Use the score to choose between the filters in front of you: sort by MK Score, then look at the top few cards' profit factor and trade count to decide which one is worth trying.
When a filter scores lower than you expected, take that as a reason to look harder before trusting it. Click its card, study the chart and the trade count, and ask yourself why the filter should work for your setup.
Compare scores within a report rather than across reports. A backtest with thousands of trades over many years gives every filter more to prove itself on than one with a few hundred.
The score on a losing backtest
The score works the same way whether your backtest makes money or not. A filter can score high while the trades it keeps still lose money.
So always check the Profit Factor on the green side of the card against the Baseline profit factor at the top of the report. The score says the filter held together. The profit factor says whether it's worth trading.
A losing backtest with a high-scoring filter that lifts it toward breakeven is a good place to start. Apply that filter and crunch again.
What the score can't tell you
The score is still measured on the same trades the filter was found in. It's a better guide than profit alone, but it's not out-of-sample testing. Before you trade a filter:
- Ask yourself why it should work for your setup. A filter you can explain is much more likely to hold up than one that just fit the data.
- Apply one filter at a time. Each filter was measured on its own, so two high scores don't add up to anything.
- Test the finished strategy on dates the Cruncher never saw, or forward test it.
Asking Claude
If you've connected Claude to the Strategy Cruncher, you can ask it about any filter's score:
Why does the top filter in this report score so low when its profit factor is so high?