Fully automated futures trading
- Thread starter globalarbtrader
- Start date
Hi GAT,
I have a question regarding the adjustment you use to compute weights given different Sharpe ratios and correlation uncertainty.
I implemented both adjustments in my system and I was surprised when I found that weights were more extreme than I was expecting in some cases. It seems that these extreme weights are produced by the Sharpe ratio adjustment, probably because I have just ~12 years of market data and the SR I measure for some strategies is very high in this period (I decided a while ago to use minute candlesticks, probably a mistake).
When I use these extreme weights my backtested SR from 1.20 to 1.55. At first I thought I was using some kind of perfect forecasting when computing the weights, but if I'm doing that I couldn't find it. Maybe by luck in this short period there was a consistency in the strategies that had good sharpe. My gut says that this increase is too good to be true.
Anyway, for now I've decided to use a hack similar to what you do with correlation uncertainty. The same way that we want to model uncertainty in future correlations, not in past correlations, I don't know how much futures SRs are going to behave as past SRs.
I was just wondering following you last reply (rolling back instrument fitting) if you had similar experience.
Thanks,
Luis
I have a question regarding the adjustment you use to compute weights given different Sharpe ratios and correlation uncertainty.
I implemented both adjustments in my system and I was surprised when I found that weights were more extreme than I was expecting in some cases. It seems that these extreme weights are produced by the Sharpe ratio adjustment, probably because I have just ~12 years of market data and the SR I measure for some strategies is very high in this period (I decided a while ago to use minute candlesticks, probably a mistake).
When I use these extreme weights my backtested SR from 1.20 to 1.55. At first I thought I was using some kind of perfect forecasting when computing the weights, but if I'm doing that I couldn't find it. Maybe by luck in this short period there was a consistency in the strategies that had good sharpe. My gut says that this increase is too good to be true.
Anyway, for now I've decided to use a hack similar to what you do with correlation uncertainty. The same way that we want to model uncertainty in future correlations, not in past correlations, I don't know how much futures SRs are going to behave as past SRs.
I was just wondering following you last reply (rolling back instrument fitting) if you had similar experience.
Thanks,
Luis
same here. ~10% daily loss. Second largest daily loss, much bigger than the ones in Mar 2020. I am a bit greedy and running vol target higher than normal. Got hammered in BZ and HO. KC/CT/CC/SB is bleeding as well. Some ZN/UB/ZB longs are helping, but not enough.
Well I spoke too soon. I guess the market sold off after my closing mark 7pm GMT, , or my mark was delayed for some reason, but I dropped another 8% when my NAV updated at 00:30 this morning :-( So I'm in line with the 10% loss seen by others.
Certainly my worst day since I started trading...
GAT
Certainly my worst day since I started trading...
GAT
Last edited:
Not significant performance, no. I'll include some stats on rules in my blog post (if I'm still trading next week!!!)
GAT
Which adjustments specifically? I've done this a few different ways. The most recent accounts for the length of data history.
"Rolling back instrument fitting" not sure what you mean here, but yes I've reverted to using simple handcrafted weights for forecast weights. My logic here was that I'd introduced so much complexity with dynamic fitting that I wanted to simplify the main part of the system a little.
GAT
Exactly, I meant the one that accounts for the length of data history. I will try the previous version of the adjustment without the bootstrap to understand what's going on, but probably I should use simple handcrafted weights for forecast weights as well.
Also started using Dynamic Optimization for live trading last week (around -10% as some others).
Thanks!
Todays trades give some insight into the DO process. I reran my backtest this morning, principally to account for the reduction in available capital, hence some of the slightly peverse buy/sell orders.
GAT
Code:
Order ID Instrument Contract Strategy Fill Time Fill Amount Fill Price
31686 DOW 20211200 dynamic_TF_carry 2021-11-29T01:02:12.000Z 1 35034
31687 NASDAQ_micro 20211200 dynamic_TF_carry 2021-11-29T01:02:40.000Z -1 16196
31688 NASDAQ_micro 20211200 dynamic_TF_carry 2021-11-29T01:08:17.000Z -1 16216.5
31689 EUROSTX 20211200 dynamic_TF_carry 2021-11-29T01:15:41.000Z -1 4150
31690 DAX 20211200 dynamic_TF_carry 2021-11-29T01:16:07.000Z -1 15399
31691 DAX 20211200 dynamic_TF_carry 2021-11-29T01:16:37.000Z -1 15397
31692 DAX 20211200 dynamic_TF_carry 2021-11-29T01:17:04.000Z -1 15394
31693 NIKKEI400 20211200 dynamic_TF_carry 2021-11-29T01:45:54.000Z -1 17800
31694 FTSECHINAA 20211200 dynamic_TF_carry 2021-11-29T03:00:10.000Z -1 15467
31695 EU-AUTO 20211200 dynamic_TF_carry 2021-11-29T08:00:43.000Z -1 640.1
31696 EU-OIL 20211200 dynamic_TF_carry 2021-11-29T08:05:13.000Z 1 269.6
31697 EU-RETAIL 20211200 dynamic_TF_carry 2021-11-29T08:05:33.000Z -1 440.1
31698 HEATOIL 20220200 dynamic_TF_carry 2021-11-29T10:29:57.000Z -1 2.1809
31699 EU-TECH 20211200 dynamic_TF_carry 2021-11-29T10:31:08.000Z -1 790
31700 EU-UTILS 20211200 dynamic_TF_carry 2021-11-29T10:31:51.000Z -1 389.1
31701 NASDAQ_micro 20211200 dynamic_TF_carry 2021-11-29T10:32:37.000Z 1 16204.75
31702 DOW 20211200 dynamic_TF_carry 2021-11-29T10:33:06.000Z -1 35056
31703 BRE 20211200_20220100 _ROLL_PSEUDO_STRATEGY 2021-11-29T10:38:40.000Z 1,-1 0.00135
Slippage
Order ID Instrument Strategy Trade Price Point Value Delay Bid/Ask Execution vs Limit vs Parent Limit Total Trading
31686 DOW dynamic_TF_carry 1 0.3749334493 -148.8485793772 -0.3749334493 1.8746672466 1.1248003479 null 1.4997337973
31687 NASDAQ_micro dynamic_TF_carry -1 1.4997337973 226.2723366602 -0.562400174 1.1248003479 0 null 0.562400174
31688 NASDAQ_micro dynamic_TF_carry -1 1.4997337973 234.5208725451 -0.562400174 23.6208073067 22.4960069588 null 23.0584071327
31689 EUROSTX dynamic_TF_carry -1 8.4847814513 632.1162181213 -4.2423907256 0 null null -4.2423907256
31690 DAX dynamic_TF_carry -1 0.8484781451 137.0292204384 -8.0605423787 0 null null -8.0605423787
31691 DAX dynamic_TF_carry -1 0.8484781451 135.7565032207 -6.787825161 -1.6969562903 null null -8.4847814513
31692 DAX dynamic_TF_carry -1 0.8484781451 134.9080250755 -7.6363033062 -2.5454344354 null null -10.1817377416
31693 NIKKEI400 dynamic_TF_carry -1 0.6625026714 359.4076992096 -1.6562566784 3.3125133568 null null 1.6562566784
31694 FTSECHINAA dynamic_TF_carry -1 0.7498668986 74.9866898625 -0.7498668986 1.4997337973 null null 0.7498668986
31695 EU-AUTO dynamic_TF_carry -1 42.4239072565 193.0287780169 -120.9081356809 4.2423907256 null null -116.6657449553
31696 EU-OIL dynamic_TF_carry 1 42.4239072565 -178.1804104771 -33.9391258052 29.6967350795 null null -4.2423907256
31697 EU-RETAIL dynamic_TF_carry -1 42.4239072565 246.0586620875 -4.2423907256 0 null null -4.2423907256
31698 HEATOIL dynamic_TF_carry -1 31494.4097422707 -6154.0076636397 -18.8966458454 37.7932916907 0 null 18.8966458454
31699 EU-TECH dynamic_TF_carry -1 42.4239072565 392.4211421223 -6.3635860885 -16.9695629026 -29.6967350795 null -23.3331489911
31700 EU-UTILS dynamic_TF_carry -1 42.4239072565 216.361927008 -4.2423907256 8.4847814513 0 null 4.2423907256
31701 NASDAQ_micro dynamic_TF_carry 1 1.4997337973 -240.5198077341 -0.562400174 1.1248003479 0 null 0.562400174
31702 DOW dynamic_TF_carry -1 0.3749334493 146.224045232 -0.3749334493 9.7482696821 8.9984027835 null 9.3733362328
31703 BRE _ROLL_PSEUDO_STRATEGY 1,-1 74986.6898625494 5.6240017397 -1.8746672466 0 -3.7493344931 null -1.8746672466GAT
Markets are somewhat recovering today. +3.5% for the day so far, but of course, the day's not done yet.