Post-mortem: 100 live trades on the 3LB Trading System

Well actually it was 101 live trades but that's neither here nor there. The number does however allow for some early and essential caveats to all else that follows. 100 (or 101) trades is a very small statistical sample and caution should be taken in extrapolating outcomes over a limited sample and limited time period. Also, what is reflected below is the result of actual live trading outcomes and as such there are trading errors, missed trades, slippage and all the other realities of a manually executed trading system in the final outcomes. (I'll try and reflect a  bit more on these elements and our plans around them later in this post)

All of that said, I do believe there are some interesting trends which have developed over the past few months and that certain findings can well assist in making some positive system optimisation improvements. But lets get into the meat of this thing:

Frequency and results
We've been live trading the 3LB system since 18 May 2015. In the period 18 May to 01 September, we've entered 101 trades which equates to roughly 1.2 trades per trading day. Of that 101 trades, we had 36 winners, 45 losers and 20 break-even trades. 

Importantly we define break-even trades as those which yield a profit or loss less than  0.2 x initial risk adopted. (So as an example if we risk 1% of capital on a trade and we lose 0.1% in the result, this is bundled with the break-even trades). To test the relevance of this threshold, aggregating all the defined break-even trade results shows that the 20 break-even trades cumulatively lost 0.4% of average capital over the 3.5 month period - acceptable as immaterial.

 So our ratio of wins to losses (excluding the break-even trades) comes in at 44.4% at an average reward-risk of 1.91-1 

The equity curve for this result is shown below:

Note that for each trade we defined our permissible risk as 1% of the monthly starting capital and the y-axis reflects the cumulative % capital aggregation. What's interesting is to note the challenges experienced mainly during the July with the 8% draw-down between 7th July and 28 July. Even with this limited sample we can start seeing and experiencing a taste of how challenging it can be to stick to one's system in such a time.

The risk and reward trade histogram for each trade is as follows :

We can see from a discipline front, when executing systems trades we've been relatively good at managing and sticking to the 1% risk rule. On the return side, the frequency of hitting the 1%, 2%, 3%+ levels is quite interesting and its something we will revisit in the optimisation section of this post.

But on balance, not the worst set of results but its within the detail that we start seeing some interesting outcomes.

Across the Trading Day
As a general rule, we've entered trades after the market open (from 09h00) up until the market close (17h00). Where there have been un-triggered setups as at the close, we've left the orders in place to trigger and play out post-market. However eye-balling our results, it became apparent that there were fairly disparate outcomes depending on when during the day a trade was opened. To analyse this further we divided the trading day into 5 distinct zones defined as follows:

• Opening (pre 10h00)
• Morning (10h00 - 13h00)
• Lunch (13h00 - 14h00)
• Afternoon (14h00 - 16h00)
• Close (post 16h00)

The outcomes after stratifying the results into these categories were as follows:

What's pretty clear even in our limited sample is that by simply not taking any trade entries before 10h00 and after 16h00, we immediately enhance both our win-rate and also positively impact our overall reward-risk ratio. What was surprising was the lunchtime results but the sample in that space was so small that this could well be an anomaly

Target versus Trail
The 3LB base system uses a step-trailing system and very often we see trades move profitably and then reverse as the steps can be quite substantial. This begged the question around whether it might not be better to set fixed targets and to forgo swinging for the fences on the huge moves by using the trailing stops in isolation. To answer this we looked back to the trade histogram (for a massive sample we could perhaps use some form of normal curve but we're really only dealing with 36 winners here so I think that's a bit premature) and noted the following return levels by winning trade:

What's clear is that the majority of the outcomes for winning trades fell into the 1% - 2% range. Given that our step-stop moves in 1% average increments  (when we hit 1% we ratchet the stop to break-even, at 2% we lock in 1%, etc, etc) this would mean that on average our trades peaked between the 2% and 3% level. So a rough optimal trade target should be at approximately the average of this range - say 2.5%. But what would the individual trade outcomes have been had we set that target level alongside the trailing stop which we still want to retain to lock in profits and preserve capital. We'd have fewer big spikes but given these compromise a relatively low proportion of overall winning trades, it should be more than compensated for by a general increase in the level of the majority of the trade outcomes. So definitely something to consider building some optimisation around.

Slippage and error
The nutshell summary is that this had a net cost of approximately 2.5% of average capital over the 101 trades. Sources ranged from trade mis-sizing to entry and exit slippage. Unacceptably high at the moment and an area of focus moving forward.

Changes for V1.1
Based on all the findings above,  the following tweaks will be incorporated into V1.1 of the 3LB system. Note that no wholesale changes can be made - its been re-iterated a few times that 100 trades does not provide a sufficient basis for that.

• No trades will be taken before 10h00 and after 16h00
• All trades taken during the "Morning" session as defined above will in addition to the step-stop include a 2.5x fixed target. The rationale here is that the impact of the US open in the afternoon session can really enhance momentum and we don't want to lose out on this by capping gains. Conversely post the open, the morning sessions tend to dwindle off so locking in quicker profits should be more optimal.
• All positions still open as at the market close (17h00) need to be closed out manually by 17h30. We don't want to hold these positions overnight and if the last trade is entered into before 16h00 it has some time and space  to move positively before the close failing which the timed close is actually ideal.
• Reducing mobile trading and looking to manually enter on more trades (as opposed to the stop orders) should reduce the slippage factor.

And that's it. Welcome any input on the above and look forward to the next 100 trades!