Tracking 100 Predictions — The Spreadsheet That Shows Whether Your Process Works
Researched and drafted with AI assistance from published sources; published by LokeNessiSport Editorial Team · Last updated: 8 August 2026
18+ | Gambling involves risk. Only bet what you can afford to lose. If gambling is causing you problems, contact BeGambleAware.org or Hjelpelinjen: 800 800 40.
Disclosure: this page contains affiliate links. If you sign up through them we may earn a commission, at no extra cost to you. It does not change what we write.
18+ only. Betting carries a real risk of losing money. Set a deposit limit before you start, and stop if it stops being fun.
The spreadsheet is not impressive. It has seven columns, no charts and no macros. It is the single most useful thing in this series, and almost nobody keeps one, because it does something nobody enjoys: it tells you the truth about your own opinions.
The seven columns
- Date / fixture: the match; filled in before
- Prediction: the outcome you called; filled in before
- Probability: your number, as a percentage; filled in before
- Mechanism: one sentence on what you expect to decide it; filled in before
- Price taken: the odds available when you decided; filled in before
- Closing price: the odds at kick-off; filled in after
- Result + mechanism check: won or lost, and whether the mechanism actually happened; filled in after
Four fields before, three after. That is the whole system. It takes about ninety seconds a match.
The three numbers it produces
1. Calibration. Group your predictions by confidence band and check the hit rate in each. Your 60% calls should win about 60% of the time; your 80% calls about 80%.
Almost nobody is calibrated at first. The typical pattern is that high-confidence calls underperform badly — you are most wrong exactly where you are most sure, which is also where people stake most. Fixing that one thing changes more than any tipping service ever will.
2. Mechanism accuracy. Of the predictions that won, in how many did your stated reason actually happen?
Advertisement · 18+ · Gambling involves risk · BeGambleAware.org
A high win rate with a low mechanism rate is the most dangerous result in the whole spreadsheet. It means you are being rewarded for reasoning that is not working, and you will keep doing it until it costs you badly.
3. Closing line value. Did you take a better price than the market closed at?
This is the professional's metric, and the reason is statistical: CLV stabilises far faster than profit does. You can beat the closing line consistently and still be down over fifty bets, because fifty is a small number. Profit over a short run is mostly noise; CLV over the same run is signal.
!Seven columns, ninety seconds a match, and it tells you the truth
What 100 rows usually reveals
Three findings show up so consistently that you can almost predict them:
Your confidence is inflated. Overconfidence is common, so do not be surprised if your 80% band wins noticeably less often than 80%. The fix is to shade every stated probability towards 50 until the data says otherwise.
Your good reads cluster. One league, one sport, or one situation — a specific kind of matchup — carries almost all of your genuine edge. Everything outside it is roughly a coin flip with a margin attached.
Volume is the enemy. People bet the most matches on the days they know the least. The spreadsheet makes that pattern visible in a way that nothing else does.
> The point of the spreadsheet is not to prove you are good. It is to find the small part of your judgement that actually is, and to stop paying for the rest.
Worked example: reading one confidence band
Suppose 25 of your 100 rows sit in the 70–79% band, averaging 75%, and 14 of them came in. A calibrated forecaster would expect about 25 × 0.75 = 18.75 hits. How surprising is 14? The spread of luck on 25 calls at 75% is √(25 × 0.75 × 0.25) ≈ 2.2 hits, so 14 is about 2.2 spreads below expectation — the kind of gap that happens by chance only around 2–3% of the time. That is good evidence your 75s are really nearer 56% (14 ÷ 25).
The fix is mechanical: until the next 100 rows say otherwise, write down about 60% when your gut says 75%, and stake accordingly. Then compare the same band again. A band with only eight calls in it cannot tell you much — the spread of luck there is almost as large as the effect you are looking for — which is why this exercise needs a hundred rows, not ten.
Advertisement · 18+ · Gambling involves risk · BeGambleAware.org
What to do with the findings
One practical footnote before you open accounts anywhere: a tracking spreadsheet plus several book logins is a lot of reused credentials in one place. Give every account its own password.
The staking side is not optional and is covered fully in sports betting bankroll management. The most common errors are catalogued in sports betting mistakes to avoid, and the value framework in how to find value bets and expected value explained.
!Track closing line value, not weekly profit. One is signal, the other is noise.
If you use tips or models instead
Run the same spreadsheet on them. A tipster or a model that cannot show calibration and CLV over a real sample is asking you to trust a highlight reel of their best month.
Our test for that is how to spot a trustworthy tipster, and the machine version has exactly the same weaknesses — see The Oracle: how AI predicts sports results.
Advertisement · 18+ · Gambling involves risk · BeGambleAware.org
The honest ending
Most people who keep this spreadsheet for 100 rows conclude that they should bet less, on fewer matches, for smaller amounts. That is not a failure of the exercise — it is the exercise working.
The margin is real, it is charged on every bet, and beating it consistently is genuinely hard. Treating this as paid entertainment with a known cost is the only sustainable version for almost everyone.
18+ only. If gambling is causing harm, stop and contact a national support line in your country.
Where the partner links sit. The "price taken" and "closing price" columns need somewhere to read a price from, and reading it from two books is better than one. We hold partner accounts with Thunderpick Sports and V.Vegas Sportsbook — partner links, both of them, and both subject to their own country eligibility terms. Nothing in this article is a recommendation to increase your stake. 18+, terms apply, your stake is at risk.
Back to Part 1.
---
*Some links are affiliate links, marked as sponsored — where a partner pays us, our analysis does not change. Gamble responsibly.*
Frequently Asked Questions
How many predictions do I need before the data means anything?
Rough calibration starts to show at around 50 and becomes reasonably readable at 100 or more. Below 30 you are mostly looking at variance, which is why short winning runs feel far more meaningful than they are.
What is closing line value?
The difference between the price you took and the price at the moment the market closed. Consistently beating the closing line suggests your estimates contain information the market had not yet absorbed, and it stabilises much faster than profit does.
Why track the reason as well as the outcome?
Because being right for a reason that never occurred reinforces a false belief. Grading the mechanism separately from the result is what stops a lucky call from being learned as a good one.
How do I know if I am overconfident?
Group your predictions by stated probability and compare each group's hit rate with its average. If 25 calls averaging 75% produced 14 hits, you expected about 19, and a shortfall that size happens by chance only a few per cent of the time, so your 75s are closer to 56%. Shade future calls towards 50% until the data catches up.