##### 2018年1月24日（水）

#### How to Assess Your Trading Performance

In this tutorial I will discuss how you can test performance of a trading system. You can also use the same techniques to test your own trading performance.

Of course the easy way is simply to look at your betting bank: If you are up then that is an indicator of good performance. On the other hand, if you just got a visit from the debt collectors who are carrying out your new HD ready widescreen TV as we speak, that is an indicator of bad performance.

But suppose you actually want to more clearly benchmark performance. In this case you will need to use something better than just checking whether your betting bank is rising or falling, as either could be due to what statisticians refer to as random variation (“luck” to you or I).

In order to eliminate luck, you need to have a statistically significant sample. Without delving too deep into the maths behind this, a statistically significant sample is where the results are so unlikely to have occurred by chance that you can almost eliminate random variation. So for example, if you are able to say that “result A” has only a 1 in 1,000 probability of occurring due to chance then that is highly statistically significant.

If you toss a coin and get “Heads” 4 times out of 4, you might be tempted to conclude that the coin is biased. But this can occur by chance (1 in 16 times). Quite low, for sure, but the sample size is very small. However, if I tossed the coin one hundred times and got 100 out of 100 heads, then I could indeed assume I was using one of Del boy’s coins as the chances of this occurring by chance are astronomic, since the sample size is much larger.

When a betting system is reviewed, such as the tennis trading league or the tennis betting system, I am not really interested in performance over the short-run. Why? Because in order to be statistically significant, I need a larger sample size.

That’s why, in my recent article on the tennis system, I stated that “ had it was not significantly significant.

With betting you also need to factor in the fact that some events have a better than even chance of occurring. This also relates to value betting, where you try identify events where the odds are higher than you would expect based on your judgement of the probability of an event happening. (But that’s another story).

So for example, if I follow a system that says “Bet on Federer to win all his matches up to the semi-final of all Grand Slams”, and I tell you my system has a 100 per cent success rate, this does not necessarily mean that my system is a good one. Why: You guessed it – this is not a statistically significant sample: In fact there is quite a high chance of this happening (if you like, it could easily happen by ‘luck’), and in order to test the system I would probably need to do so over a far greater number of years than Federer has left.

Generally speaking, if you are betting on a system where the average odds are 2.0, you would have to consistently achieve well over 50 per cent rate in order to have a worthwhile return. This is because odds of 2.0 indicate that there is a 1 in 2 chance of the outcome being successful – plus the fact that if you are using a online sportsbook rather than a betting exchange such as Betbubbles, you will offered be offered worse odds to cover their margin. If the system got 6 out of 10 that is still not very significant since it could have equally well have been 4 out of 10 or 7 out of ten due to random variation. On the other hand, if I have system that bets on 3.0 or better chances, it means on average you would only expect 1 in 3 success rate (Due to random variation), So if such a system consistently performed with an 50 per cent success rate it would be well worth looking into it.

The good news is that I take this into account when look at the system reviews that I will be undertaking over the coming weeks. If I recommend a system as a good performer, it will be far less likely to be down to simple random variation! So please look out for more reviews where you will see this principle in operation.

Similarly, when looking at your own performance, remember to look at: a) what the average odds were and b) over a sufficiently large sample before coming to firm conclusion.

This is why “return on investment” is a better measure, since it builds in the price at which you traded. Use this benchmark over a sufficiently long period and you will know precisely how well you are doing.

Of course the easy way is simply to look at your betting bank: If you are up then that is an indicator of good performance. On the other hand, if you just got a visit from the debt collectors who are carrying out your new HD ready widescreen TV as we speak, that is an indicator of bad performance.

But suppose you actually want to more clearly benchmark performance. In this case you will need to use something better than just checking whether your betting bank is rising or falling, as either could be due to what statisticians refer to as random variation (“luck” to you or I).

In order to eliminate luck, you need to have a statistically significant sample. Without delving too deep into the maths behind this, a statistically significant sample is where the results are so unlikely to have occurred by chance that you can almost eliminate random variation. So for example, if you are able to say that “result A” has only a 1 in 1,000 probability of occurring due to chance then that is highly statistically significant.

If you toss a coin and get “Heads” 4 times out of 4, you might be tempted to conclude that the coin is biased. But this can occur by chance (1 in 16 times). Quite low, for sure, but the sample size is very small. However, if I tossed the coin one hundred times and got 100 out of 100 heads, then I could indeed assume I was using one of Del boy’s coins as the chances of this occurring by chance are astronomic, since the sample size is much larger.

When a betting system is reviewed, such as the tennis trading league or the tennis betting system, I am not really interested in performance over the short-run. Why? Because in order to be statistically significant, I need a larger sample size.

That’s why, in my recent article on the tennis system, I stated that “ had it was not significantly significant.

With betting you also need to factor in the fact that some events have a better than even chance of occurring. This also relates to value betting, where you try identify events where the odds are higher than you would expect based on your judgement of the probability of an event happening. (But that’s another story).

So for example, if I follow a system that says “Bet on Federer to win all his matches up to the semi-final of all Grand Slams”, and I tell you my system has a 100 per cent success rate, this does not necessarily mean that my system is a good one. Why: You guessed it – this is not a statistically significant sample: In fact there is quite a high chance of this happening (if you like, it could easily happen by ‘luck’), and in order to test the system I would probably need to do so over a far greater number of years than Federer has left.

Generally speaking, if you are betting on a system where the average odds are 2.0, you would have to consistently achieve well over 50 per cent rate in order to have a worthwhile return. This is because odds of 2.0 indicate that there is a 1 in 2 chance of the outcome being successful – plus the fact that if you are using a online sportsbook rather than a betting exchange such as Betbubbles, you will offered be offered worse odds to cover their margin. If the system got 6 out of 10 that is still not very significant since it could have equally well have been 4 out of 10 or 7 out of ten due to random variation. On the other hand, if I have system that bets on 3.0 or better chances, it means on average you would only expect 1 in 3 success rate (Due to random variation), So if such a system consistently performed with an 50 per cent success rate it would be well worth looking into it.

The good news is that I take this into account when look at the system reviews that I will be undertaking over the coming weeks. If I recommend a system as a good performer, it will be far less likely to be down to simple random variation! So please look out for more reviews where you will see this principle in operation.

Similarly, when looking at your own performance, remember to look at: a) what the average odds were and b) over a sufficiently large sample before coming to firm conclusion.

This is why “return on investment” is a better measure, since it builds in the price at which you traded. Use this benchmark over a sufficiently long period and you will know precisely how well you are doing.

コメント(0件) | コメント欄はユーザー登録者のみに公開されます |

コメント欄はユーザー登録者のみに公開されています

ユーザー登録すると？

- ユーザーさんをお気に入りに登録してマイページからチェックしたり、ブログが投稿された時にメールで通知を受けられます。
- 自分のコメントの次に追加でコメントが入った際に、メールで通知を受けることも出来ます。