The sample-size problem
At a 55% true win rate, a twenty-ticket sample lands between eight and fifteen wins most of the time. That range contains both a broken model and a very good one, so the record itself tells you almost nothing that early.
Closing line value fixes this. If you buy at 52 and the contract closes at 57, you beat the market by five cents regardless of whether the game went your way.
How to track it
Record the price you paid and the last traded price before kickoff on every ticket. Average the difference in cents. Twenty tickets of CLV data is a far stronger signal than twenty results.
Positive average CLV with a losing record means variance. Negative average CLV with a winning record means you got lucky and should tighten the model, not add size.
The desk's rule
If rolling twenty-ticket CLV goes negative, position sizing is halved until it recovers. That happens before the P&L notices, which is the entire point of measuring it.