In the past I have been known to let bets on horses go in play if the odds have moved against me when placing the lay. This mean I have watched the odds on a lot of races. I noticed that in a lot of races horses were coming right in before going on to lose and it gave me the idea. Of course you can never guarantee which horse it will be but I decided to give it a go. It works especially well on sprints where there are lots of close races.
I don't arb the horses as much as I used to, partly because my situation at work has changed and partly because I have 6 gubbed accounts at Stans

. I always used to fully under lay my bets anyway as the 1 winner with £100 profit would cover 50 losing bets where I could have made £2 guaranteed. I have probably tried this method 15 - 20 times. In all those attempts I have had 1 horse win which cost me about £12 profit on a small qualifying bet (I still made £12 as the idea is to risk half the profit).
In the example which won I used a £20 SNR. Before the race I backed at 17 at the bookie and laid at 14 at BF. I laid to make the full £20 from the Freebie meaning that if the horse went on to win I could have made an extra £27 profit from the underlay on top of the £20 Freebie. I decided to potentially sacrifice half of this extra profit if the horse won, giving me £33.50, still V good from a £20 SNR. I then laid £135 at 1.1, a lay risk of £13.50. I didn't bother to check the liquidity at 1.1 to be honest as I knew that if the horse was heading near the end of the race it would be taken.
It is a long shot, but the reward is 10 x the profit you're risking. Question is for every winner 10 winners you back do you back one that runs well but ultimately doesn't win?