SPORTKWIQ

Track record

Every SPORTKWIQ pick, scored after the fight. Only the first prediction we published for each fight counts, saved with a timestamp before the fight, so nothing can be tidied up afterwards. Open to everyone.

9–3
picks won–lost
75%
picks correct
73%
bookmakers' favourites won (same fights)
0.651
log loss (lower is better; 0.693 = coin flip)
Method record: 5 of 12 · fights where our pick won and won the way we said was most likely (KO/TKO, submission or decision). Taken from the predictions saved before each card, so misses count too.

Against the bookmakers' closing prices on the same 11 fights: our log loss 0.668, bookmakers 0.670. We've been more accurate so far. Small samples swing a lot: judge us after 100+ fights.

Did the market move our way?

5 / 11
picks where the odds shortened after we published
+0.7
average move towards our pick (percentage points)

When the bookmakers' price moves towards a pick between us publishing it and the fight starting ("beating the closing price"), it means the betting market came round to our view. Doing that consistently is the standard sign of a real edge. 3 moved against us, 3 barely moved.

When we say 70%, does it happen 70% of the time?

Calibration: what we said against how often it happened, by chance band 0% 25% 50% 75% 100% 50% 60% 70% 80% 90% 100% perfect calibration What we said (our win chance) How often it happened 50%–55%: we said 53%, it happened 50% (2 fights) 60%–65%: we said 64%, it happened 80% (5 fights) 65%–70%: we said 65%, it happened 100% (2 fights) 70%–75%: we said 74%, it happened 100% (1 fight) 80%+: we said 86%, it happened 50% (2 fights)

Each dot is a band of picks (bigger dot, more fights). Dots on the dashed line mean our chances are honest; above it, our picks won more often than we said. With only 12 fights so far, expect big wobbles.

We saidFightsAverage chanceActually won
50%–55%253%50%
60%–65%564%80%
65%–70%265%100%
70%–75%174%100%
80%+286%50%

Card by card

DateCardRecordCorrect
2026-11-14UFC 3349–3 75%

By confidence level

LevelFightsWonExpectedActual
Very strong21 86%50%
Strong43 67%75%
Solid44 63%100%
Lean21 53%50%

Every pick

DateFightOur pickChanceResultHow it endedMethod
2026-11-14Gane vs Hokit Ciryl Gane65% ✓ Won Ciryl Gane · unanimous decision KO/TKO ✗
2026-11-14Harrison vs Nunes Kayla Harrison64% ✓ Won Kayla Harrison · rear-naked choke Dec ✗
2026-11-14Borralho vs Belgaroui Caio Borralho63% ✓ Won Caio Borralho · unanimous decision Dec ✓
2026-11-14Hasan vs Gurule Bilal Hasan90% ✗ Lost Luis Gurule · split decision KO/TKO ✗
2026-11-14Thompson vs Radtke Stephen Thompson55% ✓ Won Stephen Thompson · spinning wheel kick Dec ✗
2026-11-14Dober vs Duncan Chris Duncan64% ✓ Won Chris Duncan · unanimous decision Dec ✓
2026-11-14Miller vs McKinney Terrance McKinney65% ✓ Won Terrance McKinney · punches Dec ✗
2026-11-14Johnson vs Almeida Donte Johnson65% ✗ Lost César Almeida · punches Dec ✗
2026-11-14Gordon vs Costa Melquizael Costa82% ✓ Won Melquizael Costa · unanimous decision Dec ✓
2026-11-14Chiasson vs Mesquita Beatriz Mesquita74% ✓ Won Beatriz Mesquita · rear-naked choke Sub ✓
2026-11-14Heili vs Díaz Juan Díaz63% ✓ Won Juan Díaz · unanimous decision Dec ✓
2026-11-14Elder vs Sanches Samuel Sanches51% ✗ Lost Evan Elder · unanimous decision Dec ✗

Underdog of the week

Each card, the underdog our full analysis gives the best chance of an upset. Higher risk, and scored separately from our main picks.

0–1
DateUnderdogBest priceResult
2026-11-21Nikolas Motta over Fan 2.30 Waiting
2026-11-14Charles Radtke over Thompson 2.33 ✗ Lost

Tested blind on past fights

Separate from the live record above. For each year, our stats model was trained only on fights from before that year, then asked to pick every fight in it without knowing the winner. 8,399 UFC fights since 2005: 60.8% correct. This is the stats model alone; our published picks also use bookmakers' prices.

YearFightsCorrectAccuracy bar
2026488 60.5%
2025513 63.3%
2024513 62.6%
2023504 61.7%
2022506 61.3%
2021497 59.2%
2020444 65.1%
2019506 55.7%
2018469 59.5%
2017446 60.5%
2016482 59.8%
2015464 61.9%
2014493 59.2%
2013376 60.1%
2012333 59.2%
2011295 59.3%
2010250 58.8%
2009213 61.0%
2008201 62.2%
2007169 67.5%
2006158 63.3%
200579 65.8%

Bars start at 50% (a coin flip) and fill at 75%. Model version 0.3-model-plus-market, tested 2026-11-11.

Draws, no contests and cancelled fights aren't counted. Past results don't guarantee future ones.