How Greyhound Trap Bias Actually Works

"Always back trap 1." It's the oldest piece of greyhound-betting folklore, and like most folklore it's half-right and half-misleading. Trap bias exists; it varies dramatically by track; and it's smaller than punters think it is. This article shows what the data actually says using our 180-day track × trap rolling stats and explains how the model handles it.

What "trap bias" means

A track has trap bias if, all else equal, dogs in some traps win more often than dogs in others. The classic intuition: at a tight track with a sharp first bend, inside traps (1 and 2) get to the rail first and benefit from the shortest line through the bend; outside traps (5 and 6) have to come wider, lose ground, and finish behind.

At a galloping track with a longer run-up to a sweeping first bend, the effect can almost vanish — every dog has time to find a line.

The numbers — where bias is largest

Below are illustrative figures using TrapStats' rolling 180-day track × trap win rate (the same data the model uses as a feature):

| Track | Trap 1 win % | Trap 6 win % | Sample size | |---|---|---|---| | Romford | ~22% | ~12% | ~4 800 entries | | Hove | ~19% | ~13% | ~3 800 | | Towcester | ~17% | ~14% | ~4 100 | | Newcastle | ~18% | ~15% | ~3 300 | | Kinsley | ~16% | ~15% | ~3 300 |

(Numbers are approximate and update nightly; pull the live figure from /api/v1/tracks-seo/{slug} to see today's value.)

Two things to notice:

1. The effect is real but bounded. Trap 1 typically wins 5–10pp more than trap 6. That's a bias, not a guarantee.
2. The gap varies by track. Romford is the steepest (10pp); some flatter tracks are nearly bias-free.

Why the bias is small in absolute terms

Take Romford at its extreme: trap 1 wins ~22% of the time vs trap 6 at ~12%. The random expectation in a 6-runner field is 16.7%. So trap 1 is about 30% over baseline, trap 6 about 30% under. That sounds enormous.

But the bookmaker prices reflect this exactly. Inside-trap dogs at Romford get short prices precisely because the market knows they win more. The bias is public information — not an edge a punter can exploit by always backing trap 1, because the price already absorbs it.

The edge appears only where a specific dog has been underpriced for its trap-track-combo by the market — which is what the model tries to detect.

How the model uses trap bias

In the LightGBM predictor, we include three features:

• track_trap_win_rate — the dog's trap × this-track rolling 180d win rate.
• track_trap_place_rate — same but for place.
• track_trap_sample_size — how many entries went into that average.

The sample size matters: a 24% win rate for trap 4 at a small track on 80 entries is much noisier than 22% at Romford on 4 800 entries. The model uses the sample size as a confidence weight automatically — leaning on the bias signal where it's well-evidenced and ignoring it where it isn't.

What changes the bias (and what doesn't)

Things that move trap bias:

• Track maintenance. A re-laid sand surface in spring can flatten or reverse some trap biases for weeks until the surface settles. Trainers know this; the data picks it up after a couple of weeks.
• Box draws vs starting traps. Some HP (handicap) races stagger starts, partially neutralizing trap bias by giving outer traps a yardage advantage.
• Specific weather. Heavy rain that softens the sand changes how dogs find pace and lane — early-pace dogs in inside traps benefit even more on softer going at sharp tracks.

Things that don't:

• Punter folklore. "Outside is best on full moons" is not a thing. The data has been measured for decades; there's no hidden second-order pattern punter chat hasn't already turned into the prices.

Practical takeaways

1. Don't always back trap 1. The price already reflects the average bias. You'd lose money over time.
2. Do consider trap × dog fit. A pacy frontrunner in trap 1 at Romford is the strongest possible structural fit. A slow-starting closer in trap 6 at Romford is the worst.
3. Look at recent track win rates, not just historical. Surfaces drift. Six months of bias may not be the bias today.
4. Use the racecard's running-style notes. "Led to bend 1" + inside trap at a sharp track = predictable; "ran on at the finish" + outside trap = often disadvantaged at the same track.

For most punters, trap bias is a useful frame but not an edge you can trade alone. The model bakes it in as one feature among many — and on the track hub pages you can see the current bias numbers for any UK or Irish stadium we track.