The claim
Pascal has read our page on the low draw, and he has a reply ready. Everyone knows about the inside rail, he says, so of course the low stalls are overbet. The market has heard the story too many times. The clever money must therefore sit at the other end of the stalls. The high draw is the unfashionable side, the one the course guides warn you about, and an unfashionable side should be the one offering a fair price.
He has a few courses in mind to back him up. On some straight sprint tracks the stands' side rail, often the high side, rides faster when the ground is drying or the wind is behind. At some round courses the bend runs the other way, and the outside stalls get the better run. Even where the draw does nothing at all, Pascal argues, a horse drawn wide at least avoids getting boxed in on the rail. Back the highest-drawn runner in every Flat race, he says, and you collect all those advantages at a price nobody else is chasing.
There's real thinking here. Draw bias does exist, it isn't always low, and contrarian betting has a respectable history: if the crowd overpays for one story, the opposite story can be underpriced. Pascal has taken the lesson from our low-draw page seriously and turned it round, which is exactly what a curious punter should do.
So we tested the mirror image. Our low-draw test lives at /ai-race-predictor/lab/does-a-low-draw-win-races/. This page takes every Flat race in our database with draw data, 20,141 races, and backs the runner in the highest-numbered stall in each one, to the same flat stake at Starting Price. Then we added it up.
Why people believe it
The contrarian case has a lot going for it on paper. Racing is full of stories about the inside rail, and a story that well known should be priced. If thousands of punters lean towards the low stalls every time a course guide mentions the bend, the horses drawn wide should drift a little longer than they deserve. Pascal's instinct to oppose the crowd is the same instinct behind most serious betting thinking.
Then there are the courses where the high draw really is the place to be. Straight-course sprints often split into two groups, and on some days the group racing nearest one rail is clearly quicker. Watch a big-field sprint where the far side comes home well clear of the near side and you start to believe the draw decided it. When that quick side is the high side, the high stalls look like a gift.
There's a tactical argument as well. A low-drawn horse can be trapped against the rail with nowhere to go, and plenty of punters have watched their selection stuck behind a wall of horses while the winner swept round the outside. The horse drawn wide has a clear run, even if it covers a little more ground. To a punter tired of hard-luck stories, that freedom looks like value.
The flaw is that each of these advantages only exists at certain courses, over certain trips, on certain days. A system that backs the highest stall everywhere buys the high draw at every track, including the many where it means nothing, and takes whatever price the horse in that stall happens to be.
Where the money goes
The highest stall is mostly just a stall. At most tracks, over most trips, the draw makes little or no difference to the result, and the horse leaving the top stall is simply whichever horse the ballot put there. Most of the time that is not the favourite. Very often it is a horse at a long price. So for most of its bets this system is close to picking a runner at random, and blind price-taking pays the bookmaker's full margin with no judgement to set against it.
Pascal's contrarian point does earn something, and the next section shows how much. If the low stalls carry a premium for their famous story, the horses drawn wide are spared that premium, and the high draw should lose less than the low draw. That part of his reasoning holds up. What it can't do is turn an ordinary price into a generous one. Being spared an overcharge is not the same as being offered a bargain.
The places where the high draw really helps are known places. When a straight-course sprint is riding fast on the stands' side, the course reports say so, the jockeys know it, and the market moves before the race. The horses drawn on the favoured side shorten in the betting and the others drift. At the tracks where the high side matters most, it is anything but overlooked.
And the margin, the overround, comes off every bet. Add up the chances implied by every price in a race and they come to more than a hundred per cent, and longer-priced horses tend to carry more of that surplus than favourites do. A system that drifts towards outsiders pays the heavier end of it, however clever the reason for picking them.

How we tested it
We mirrored the low-draw test exactly, so this page and that one can be read side by side. The test covers Flat races only, because jump races have no starting stalls, and only races where a draw is recorded in our database. In each race we found the runner leaving the highest-numbered stall among the horses that actually ran, and backed it with one flat stake at the official industry Starting Price. That gives 20,141 races out of the 31,308 real British races we hold.
Non-runners are voided, and that matters on a draw page. If the horse in the top stall is withdrawn, the highest stall that actually ran becomes the selection, which is how the draw plays out on the day. A race needs a recorded winner to count, and if the top recorded draw is ever shared, the stake is split between the runners sharing it, as it is for joint-favourites.
Starting Price is the generous figure. It assumes the official return with no commission and no shading, so a punter taking real prices would usually do a little worse than the number in the next section.
We used no course filter, no distance filter and no field-size filter, on purpose. Pascal's theory is that the high side is the unfashionable side and is underpriced across the board, so the blanket test is the fair test of it. Whether the high draw pays at the particular sprint tracks where it is known to help is a narrower claim, and it would need a page of its own.
The selection is the stall, not the horse. The highest-drawn runner might be the favourite or a big outsider, and the test takes whichever it is.
The numbers
Here is the number. Backing the highest stall in every Flat race with a recorded draw returns -16.41% across 20,141 races, so about £83.59 of every £100 staked comes back over the long run. The top stall wins 12% of these races, about one in eight, and the 95% range on the result is [-21.3,-10.9]. Both ends of that range sit well below zero, so this is a heavy loss whichever way the noise falls.
Now the comparison Pascal was waiting for. The lowest stall, tested on exactly the same races in exactly the same way, returns -25.00%. A runner picked completely at random returns -21.42%. On the headline figures the high draw beats both, and by a fair margin over the low draw. The two draw tests win almost exactly as often, so the difference is in the prices: the horses in the top stall have been paying out at better odds for the same strike rate.
That fits Pascal's theory more closely than most theories on this board fit anything. If the low-draw story makes the inside stalls a little too popular, the opposite end should be a little less overpriced, and that is what the numbers show. We'd still hold the claim loosely. The ranges around the three figures only just touch at their edges, so the ordering looks real, but the precise size of the gap is uncertain.
What the comparison can't show is a profit. The high draw loses less than the low draw and less than a pin, and it still loses heavily: far more than the plain favourite, which returns -9.03%. Escaping one overcharge left the bet facing the ordinary margin on a lot of long-priced horses. These figures pull live from our data, so the ranges will tighten as more Flat seasons come in.
The verdict
A high draw doesn't win races often enough to pay. Backing the highest stall in 20,141 Flat races returns -16.4%, about £83.59 back for every £100, with a 95% range of [-21.3,-10.9]. That is a heavy loss, and no blanket draw rule we have tested gets anywhere near break-even.
Pascal deserves real credit, though. He read our low-draw page, spotted that a famous story might be overpriced, and predicted that the other end of the stalls would do better. It did. The high draw loses noticeably less than the low draw, -25.0%, and less than picking a horse at random, -21.4%. That is a contrarian idea making a testable prediction and passing the test, which is rare on this board.
The part that fails is the jump from "less overpriced" to "underpriced". Most of the time the horse in the top stall is an ordinary runner at a long price, and long prices carry the heaviest share of the bookmaker's margin. Where the high side really does ride faster, the market sees it on the day and adjusts. Escaping the low-draw premium shrank the loss without coming close to removing it.
The useful lesson travels well. When a story is famous, its price usually includes the fame, and betting against it can lose less. Losing less is still losing, and a draw is a fact about a stall, not a judgement about a horse.
If you enjoy a draw angle on a sprint afternoon, keep the stake small and treat it as part of the fun, never money you need. Past performance is no promise of future returns, and our Starting Price settlement is kinder than the prices most punters get.


