Losing strategies wanted
I once found a fantastic way to make money. It was so simple. I started following Brimardon and found that I was losing money fast backing there losers. When I finally ran out of money I backtracked having kept records and found that if I had layed every winning selection I would have made a fabulous profit...trouble was I daren't do it in case they actually picked more winners!
I'm currently evaluating a couple of systems like that... All against historic data (so 'what would have happened if..', rather than live trading)freddy wrote:Backing and laying simultaneously at the current back and lay odds is a good way to lose money
Back all horses when their odds are within range 'A' to 'B'
Immediately Lay at 'C%' less than the Back Odds
Lay trade value is determined so that it would create net Profit of Back value + 'D%'
Only trade when race is 'E' seconds into in-play.
Use 'F' Back odds (e.g. Low, Medium, High)
A, B, C%, D%, E and F can each have multiple values. So re-run the calculation for each defined permutation. The values I selected give just under 6,000 variations.
So one real example:
Back all horses with odds between 10 and 26
Lay at 25% less (so, if back is 10, lay at 7.5)
Lay value is set, based on what would be required to make a Net Profit, assuming horse loses, of Back value + 200%
Only start trading 120 seconds into in-play.
Back using 'Middle' odds.
My simulation then runs through historic data, horse by horse, race by race. If a 'back trade' opportunity occurs (e.g. between 10 and 26), then look to see if the LAY Odds fall between that point in time, and the end of the race. If they do then it assumes the lay is made. Available liquidity is respected so not all trades are the desired value. Finally, check the actual race winner to determine what the event/horse P&L is.
The second variation on this theme just works with a Lay then Back (so all in reverse).
The most astonishing thing is just how bad both systems are. The system is still chundering through the permutations, but as it stands, it's calculated the result for 27,117 days (well, about 10 days of races, using just over 2,700 permutations. Out of those 27,000 'days', it lost money on all but 69 days.
I'm considering writing a new system that just bets on everything that this system doesn't bet on. On the basis that this was so awful, then the complete opposite might be good.
I thought this approach up whilst trying to sleep one night. I had to get up, go to my office, write up the outline idea before I could go back to sleep. That's 30 wasted minutes of my life that I'll never get back.
Having said that, writing the code to test the idea against historic data was, as always, interesting.
Incidentally, I do have better systems, including one I'm trading live at the moment. So please don't throw me out of the BetAngel trading club for having the worst system ever.
Im glad someone else is like that! When I have a long car trip, I switch the radio off and think through new strategies.Ive come up with some of my best ideas at times like that!EyePeaSea wrote: I thought this approach up whilst trying to sleep one night. I had to get up, go to my office, write up the outline idea before I could go back to sleep. That's 30 wasted minutes of my life that I'll never get back.
:
think of it as 30 mins well spent, as you have just crossed one off the list and you're now one nearer !
I'm re-reading this interesting thread as I'm considering creating a scalping bot after years of trying to profit by betting at 'value' prices. This remark really struck me. Just to clarify, are you saying that any automated strategy whose first step is to take the best price on offer will automatically lose in the long run?Euler wrote:If you deploy strategies that take prices they will always lose in the long term because the market is pretty efficient and you will lose money on the spread.
Very interesting question.Wyndon wrote:Just to clarify, are you saying that any automated strategy whose first step is to take the best price on offer will automatically lose in the long run?
It's widely assumed that you do worse by taking from the market than offering, but I'd be interested in knowing if anyone has done any studies to test that hypothesis.
Jeff
Euler wrote:If you deploy strategies that take prices they will always lose in the long term because the market is pretty efficient and you will lose money on the spread.
That's all slightly over generalised. Any random "strategy" will lose you money. But if you can identify value, then it's possible to make money (i.e. you are no longer betting at random, but selectively).Wyndon wrote:Just to clarify, are you saying that any automated strategy whose first step is to take the best price on offer will automatically lose in the long run?
I do have strategies that take prices which are profitable. In fact one of my longest running strategies is a "take" strategy. It's not as profitable as it used to be though, which does suggest markets are more efficient. The strategy obviously has to have a big enough edge in order to make a profit, but it's not impossible, so I wouldn't suggest ever assuming something is true.
I do have a similar strategy that offers money, and that too is profitable. Offering money is a lot harder to backtest though, as you can never know if someone would have taken your money if it were there waiting on the order book.
I suspect it's being a market expert that allows you to detect a subtle quirk. For most people it wouldn't even be noticed, let alone be acted uponEuler wrote:Curiously, I did eventually find a consistently losing strategy. But I'm not in a position to share it unfortunately. But it was found by accident and was very un-obvious. A subtle quirk in the market it seems.
Very interesting point.xitian wrote:I do have a similar strategy that offers money, and that too is profitable. Offering money is a lot harder to backtest though, as you can never know if someone would have taken your money if it were there waiting on the order book.
One way of testing offering vs taking would be to offer and take at random using £2 stakes, with stops of x and offsets of y, and see which leads to the least loss.
Something that would need to be taken into account would be the opportunity cost, i.e. the times when, had I taken from the queue rather than offered and not got filled, my offset would have been hit. That is, you may be missing out on lots of successful trades by waiting to be filled. With a random system, that means you will lose less money, but if you have an edge it could mean that you will make less money.
Jeff
Let's assume that, in this random market, I have an edge. Let's also assume that the fill rate is low.
If, by taking, I can make two trades in the time someone else makes one trade, then my overall profit may be higher than theirs (even though their profit per trade is higher). Also, if they are offering to close, then they may take bigger losses than I do by taking to close.
However, theoretically, in a truly random market, it doesn't matter what you do - it all evens itself out, and the expected long-term negative edge is the same. It seems to me that you only get an edge through offering in a random market if you're up against enough 'impatient traders' who create inefficient liquidity by entering at non-optimal places.
However, in a market that's not random but is moving with gusto in a particular direction, if I offer in the direction of the market then few people will be interested in accepting my offer. If a market is rapidly drifting, and I offer at 6.0 (along with everyone else who wants to ride the move), I might miss the boat if the market doesn't stop to catch its breath till it reaches 6.4-6.6.
Jeff
If, by taking, I can make two trades in the time someone else makes one trade, then my overall profit may be higher than theirs (even though their profit per trade is higher). Also, if they are offering to close, then they may take bigger losses than I do by taking to close.
However, theoretically, in a truly random market, it doesn't matter what you do - it all evens itself out, and the expected long-term negative edge is the same. It seems to me that you only get an edge through offering in a random market if you're up against enough 'impatient traders' who create inefficient liquidity by entering at non-optimal places.
However, in a market that's not random but is moving with gusto in a particular direction, if I offer in the direction of the market then few people will be interested in accepting my offer. If a market is rapidly drifting, and I offer at 6.0 (along with everyone else who wants to ride the move), I might miss the boat if the market doesn't stop to catch its breath till it reaches 6.4-6.6.
Jeff
Euler wrote:On this particular point, imagine the market is random and moves from one random point to another. How do you expect taking to be better than offering?
Thanks for all the responses. I feel that this could form the basis of a first law of trading. "Taking the best price on offer on a random basis will always lead to a loss in the long run". This may prevent the loss of an awful lot of time and money for people like me who thought that with a bit of 'clever' manipulation the loss inherent in the spread could be overcome. Does anyone know any other axioms which could usefully be employed in designing a strategy?