VicIF Exploration - Part 2

In the first part, I get to a bunch of related stats that the strategy could be adjusted upon and figured out how to run custom csv backtest after manipulating the csv files.

From there, I noticed one of the most important thing I should look into is to automate the generation of the custom csv. As I showed that I spend like 1hour doing the csv manipulation just for that test. So I wrote a simple python program just to do that, take in the backtest tradelogs and convert it into the custom csv.

That's pretty much how it looks like, simple. It takes in 2 tradelogs csv because I run put and call separately then download each of the tradelogs which result me in 2 files. So instead of manually combining them, I also get the program to process both together and generate the custom CSV output file that is required for the custom backtest.

With this, each variation backtest run can be achieved in minutes, so I ran a bunch of them. Before I go into the numbers, you'd also realise there's a summarise button there. Basically I also made it summarise all the stats for me instead of me doing the excel work.

This picture really has alot of very important stats. Just to explain alittle, all these runs are 10 wide, changes in entry delta, entry credit target or tested offset. Let me just talk about a few of them to better understand.

25 Delta 15 Tested

Let's take the most right side one for example, it's 25 delta +15 tested. Means the entry will be on 25 delta and the IF conversion trigger will be when SPX moves within the strike +/-15pts. E.g. if it's a put spread at 3800 short strike, when SPX moves to 3815 it's considered tested and the IF is converted.

So on this 25 delta 15 tested variant, we see average entry premium around 428.8 and average ironfly conversion premium at 591. This means we collected > 1000 for the initial premium plus the conversion. Which is why you will see on the single stop P&L stats, it's actually slight positive.

This variant basically result in a lot of iron fly conversion and is kinda betting on the fly pin profit to outweigh the losses. Apparently from the backtest it does seem to work but we cannot just read numbers, we need to interpret what you're expecting from here. It means we are going to go through bigger drawdowns and bigger wins, hence a much larger roller coaster ride in the strategy itself. With the percentage of single stop and double stop you can imagine, you're basically always stopping out converting IF and hoping it pins a good profit.

15 Delta 5 Tested

So on the most left, is 15 delta 5 tested. This is pretty much the opposite of the one above. With a much lower entry credit of 250.88, together with the IF conversion credit of 473.54. We only have about 724 out of the 1000 (10-wide), means we leave 270+ on the table. This 270 is also the losing amount if we had single stop.

Which is where you see there's a much larger loss on single stop P&L but because it's a much lower delta, the double win % is pretty high. Effectively breaking even the single stop amount. Double stop is also minimal here leaving the fly pin profit as the edge on this variant.

So from the percentage, you can imagine in this variant you'll always be getting double win or single stop. Then for those stops which converts to an IF, you get some bonus out of it. One of the very big difference from this variant to the 25 delta one is there are significant lesser amount of IF created. Which is also why there's smaller IF profit.

So from here I'm already telling myself somewhere in the middle might have a good balance of all these stats. Which leads to 20 delta.

20 Delta 10 Tested // $5 10 Tested

This 2 variant is pretty similar just that the way entry is determined is different. 20 delta in most cases are near $5-5.5 in the short leg credit. So after doing the 20 delta one, I tried running the $5 one to see if credit target gets a more consistent result.

We can see the entry credit and IF credit are pretty similar, 334-338 and 544-548 respectively. This leaves about ~120 on the table, which is significantly reducing the single stop loss which is seen on the 15 delta variant. Although we had alot more single stop as shown in the percentage, each loss is much smaller.

But what's interesting is, by losing alot more times with smaller amount, we generated more flies than the 15 delta one which increases the pin amount by quite a big percentage. Almost as many pin as we would get in the 25 delta variant. In short, we're converting cheaper flies.

Hence we are getting a much bigger profit from the fly pin which made this variant statistically the better one in backtest. Here's the OO backtest on this variant:

Conclusion

Disclaimer first, all these are backtest result done on OptionOmega. I've yet to trade any of these live. These advance way of running backtest on OO opens up alot more possibilities but it does require some additional effort and technical skills to make things easy. Hope OO can provide more tools to make these custom backtest easier.

Along the journey running these backtests, I can totally see which stats needs improving and try to tweak and run to see if I can balance the stats. From the different variant of VicIF, the last one seems to be the most balanced to me and yet the highest expectancy from both backtest result and summarise statistics.

Note that all these are tested with 9:33am entry on both side. Different timing result may vary quite alot. Which leads to further exploration that can involve different timing entries and possibly multiple tranche entries within the day.

Personally this is more of a backtest project, I wouldn't be trading this any time soon as I have other plans on my trade allocation. Let me know if there are enough interests in this, then I'll do a part 3 backtest that involves multiple tranches and different timing entry.