There’s a general understanding that if you manage to gather a high amount of profitable data, then this automatically means you’ve discovered a winning system. For example, if you had managed to gather data for 500 tennis matches which fitted your specific match conditions criteria of backing specific big servers on fast surfaces, and a good typical profit target had been achieved over these 500 results, then that would be enough to believe you had found a winning system and to begin trading it with full stakes.
But just how confident can you be that you have indeed discovered a winning system that will continue to show profit beyond your initial gathered data? You may be surprised with the answer. Read on to save yourself from a potentially fatal error.
First of all it’s important to know just how very realistic it is that a freak batch of results may have occurred within the profitable data you’re basing your system on. Using the above tennis system example, if we say that the average price of these 500 trades was around 2.00, it would only take about 12 of these winning results to have been losers for the entire profit to have been wiped out. Or we could say, if someone started the system after those initial 500 profitable results, it would only take an extra 24 losers over winners to mean the system in it’s historic entirety was now no longer in profit. And these 24 losers would not even have to fall in some kind of a consecutive streak.
And just how realistic is it for some kind of a consecutive streak of freak results to occur? Let me share with you some of my own experiences to help give you a feel. On a tennis system I was once testing, within only the first 100 results I had already managed to experience a winning streak of 19 consecutive winners at an average price of 1.66. The chance of a 1.66 trade winning 21 times in a row is only
1 in 15,207. (The calculation for this is 1.66 x 1.66 19 times). A more astonishing example than this is when two friends and I were in the snooker club and we had to toss a coin to see who was the odd one out so that the first match up could be determined. Well, between us we tossed a tail an incredible 23 times in a row! We had all gotten quite hysterical by the end of it and once that head eventually came up we all sprinted to the counter to get our hands on a calculator to work out the chance of what had just happened. The calculator showed that the chance of tossing tails 23 times in a row is an extraordinary 1 in 8,388,608!
And it doesn’t end there, during my 18 years of sports trading I have consistently experienced many other highly unlikely winning and losing streaks which were either consecutive like the ones above or which included the odd opposite result thrown in between. As you can see the chance of each of these kinds of streaks occurring are mathematically very unlikely, but trust me they certainly do pop up from time to time and they could happen to anyone and with any system.
So what does this all mean? It means that if you’re relying heavily on data to be confident on your system, freak streaks of results occur far too easily to be rely merely on 500 or even 1,000 results of data. Remember that within a batch of results, a freak occurrence of data does not even need to occur in any particular order or consecutively like in the two personal examples I shared above. Therefore the chance of freak data occurring simply anywhere within a batch of data is far more likely, and probably a lot more likely than what one would think.
So couldn’t you just wait a bit longer to gather a larger amount of results to improve your confidence in the data? For example wait for 2,000 or 3,000 results? Yes you could, but there are usually a number of serious problems with doing this. One is that in your testing criteria may only occur enough times that you would have to wait an enormous amount of time, maybe years, to gather enough data. Another problem is that data loses its value over time because the market and sports are always evolving, and so for how long a period of time is your data really going to be relevant enough?
Okay so this all sounds rather difficult. How is someone meant to find a winning system which they can be confident enough on with just a limited number of results? The answer is to combine your data with a second factor. Logic.
Using your logic to help justify why your system should prove profitable in the long run is a key element to your success and is absolutely crucial during many cases where you will not be able to gather an enormous amount of relevant enough profitable data. The most important thing to remember when applying logic is to ask yourself ‘Why should the prices of this system be undervalued in the market’? To help answer this question to yourself, it’s important to have a good understanding of the market and what other people are thinking and doing to produce all the market liquidity which your system’s trades will effectively be competing against.
If it’s likely that lot’s of other traders are already aware of the information which your system is based on, then it’s also likely that the market is already reflecting the system’s information in it’s price and that there’s not enough value left in the market price for the system to still prove profitable. On some occasions the information may be so widely known and focused on that it may actually be over valued by the market and potentially throw up a winning system in the opposite direction of what you initially thought! Be open minded to this as you may occasionally stumble upon a system which you really didn’t expect!
On the other hand, if you think you’ve discovered a system which very few other traders are aware of, that’s a good sign that you may have found a true winning system containing undervalued trades. It’s also helpful to have an understanding of where most of the market liquidity is coming from. For example, if you think that a lot of the market’s liquidity is coming from scalpers who are only really interested in getting a large amount of money matched on both sides of the market without having much interest in whether the market is actually being pitched in the right place or not, then this offers other value seeking traders a better chance of capitalizing on systems which look to expose when the market is currently being pitched too high or too low.
In summary, by combining a large as possible amount of relevant profitable data, together with sound logic as to why your system should logically be producing trades which are being undervalued by the market, you will be giving yourself the best chance possible that your system will continue to produce profit it the long run.
It’s impossible to say exactly how much merit to give to data in comparison logic, as it largely depends on the strength you feel you have of each. But as a general rule I would suggest that applying two thirds merit to your system’s data and one third merit to your system’s logic will be a sensible balance in many situations.