Not fully related to this, just a thought: there's probably 2 possible ways of looking at trading strategies and why is it even possible to profit: first is "inefficiencies", and the second is like "competing organisms in the jungle". (there's also the compensation for risk one but let's forget about it for a second). If it's the first one, then it follows that when we (and I mean "we" as all people who actively trade (I recall a funny moment in one of the videos when Jim S. asked to clarify "what do you actually mean by "we" ? " when a guy from the audience who was sort of referring to both of them as "we" in his question ) ) trade and continue to improve our strategies, we will eventually drive out all of these inefficiencies, and basically there will be nothing left to collect. Probably even including the risk-premium stuff, although, there are certain limits on these things e.g. not all market players can take all the risk because of their investment-mandate or risk-tolerance. If it's the second, then I guess it's a more optimistic situation for active traders, because it implies that there's no "proper equilibrium" that can be achieved, and it's always like "you push it from one side and something else pokes out from the other, which someone can exploit". Like e.g. if in the jungle one species become too successful\dominant, this will only shift the reality such that it will open opportunity for another spices to exploit it. So it's like a never-ending tug of war "you do this, then I'll do this and make a profit", and "if you fix that hole, it will open another and I'll come from another end and find another way to make the same profit"..
I think all of these explanations are feasible, and co-exist in different areas of the investing space. So it's completely plausible to me that HFT nowadays is more like competing organisms in the jungle, than picking up inefficiences. It's equally plausible that (for example) large institutions rebalancing their books create inefficiencies in intra-month trading patterns which will eventually be competed away. It strikes me that most markets will migrate from ineffeciences towards jungles, as more predators come in and eat up the remaining 'civillians', leaving only themselves to fight against. Risk premia, which is where I think trend following and carry sits, is more interesting. We can view it as a type of particularly vague inefficiency, with low SR, a lot of cyclicality in returns, and therefore an unattractive prospect rather than a rich and rewarding inefficiency which will be competed away. Or more interesting, if we take the view that not all market participants value risk the same, then risk premia will continue to be paid from natural sellers of risk to natural buyers. GAT
It's not a secret and there are no barriers to trade it. So it's not alpha, per se (though there can be alpha in the implementation details) It's relatively low SR, although can be reasonable once diversified across many markets (there are barriers to diversification, as more capital is needed, and too much capital is also problematic) It's performance is cyclical, as risk appetite wanes and waxes People don't like holding trend following portfolios as they violate natural human instincts to take profits early and let losses run. They are willing to pay a (risk) premium to avoid feeling uncomfortable. More often that not you lose money, which again is uncomfortable. GAT
when I hear the term “carry trade” I think of something that earns you every day until it blows you up (like owning a bond or selling volatility). You are implementing it the other way.
No, that's not right. The attributes of a carry trade depend a lot on the underlying asset class: selling options: heavily negative skew FX: fairly negative skew, especially EM commodity futures: modestly negative skew bond futures: modestly negative skew (positive in the US) equities: modestly positive skew credit: very positive skew (from table II in https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2298565) I'm doing a mixture of 2-5, with a bit of 1 from my structurally short VIX/V2X positions. So overall I should have negative skew, although so bad that I will blow up one day. GAT
Interesting.. What I also meant by "competing organisms in the jungle" is not necessarily predator-pray relationship, but more like a waterbed type of situation, i.e. for example at some point carry premium might become more attractive, so people will start trading it more, which will drive it's premium down, but maybe at the same time this will make some other strategies e.g. trend more attractive as a consequence.. So by capturing return of one strategy maybe traders inevitably make other strategies more profitable and the whole thing always exists in this "chaotic state" where some opportunities always exist. Unless everyone just trades all possible strategies in "equal proportions" at all times.. Interesting to think of risk premia as a vague inefficiency. I mean if we "know" that by taking these risks we "will definitely" make money (maybe slowly, maybe not steadily), then it becomes an obviously-rational thing to do, so it's not really risk anymore, because we "know what will happen". But I guess it's just wrong to think in these terms because risk by definition is not a certainty, it's not a guaranteed-to-happen thing, no matter how much it's studied and researched.. And + of course the prospect theory stuff that "your personal situation matters", so the risk-appetites of a middle-class family, a pension fund and a person approaching retirement are different.. I've also heard of this idea of "priced risk" vs "not priced risk". I.e. there are lots of risks in the market, but not all of them are compensated. e.g. equity vs bond premium is widely considered priced, but an idiosyncratic risk of holding only one company isn't.. Or say if my random strategy is "always short wheat", it it risky ? - yes, but is there any reason to believe that I will be compensated for this risk - no. (unless maybe I'm doing it because I'm betting on an unusually-high supply this year because of the weather or something, then maybe this risk will become "priced")
Question for Rob (or anybody else using his method of forecasting), how often should one revisit forecast scalars? I've been using MAC, breakout and carry rules, based on both of your books, scaling factors for MAC are from Systematic Trading and for breakout from Leveraged Trading and I can't seem to hit an average of 10 in the backtest. I tested on 35 instruments, data since Jan 1st, 2001. I computed an average by computing the rule forecast for each instrument, and then Code: statistics.mean(forecasts_df.abs().mean()) The `abs().mean()` gives an average for each instrument, then `statistics.mean()` gives an average across all instruments for a given rule. Here are the values I get: Code: # rules and forecast scalars rules = { 'mac_2_8': 10.6, 'mac_4_16': 7.5, 'mac_8_32': 5.3, 'mac_16_64': 3.75, 'mac_32_128': 2.65, 'mac_64_256': 1.87, 'breakout_10': 28.6, 'breakout_20': 31.6, 'breakout_40': 32.7, 'breakout_80': 33.5, 'breakout_160': 33.5, 'breakout_320': 33.5, 'carry': 30 } for rule, scaling in rules.items(): print("Avg. forecast for {}: {}".format(rule, average_forecast(all_symbols, rule, scaling))) Avg. forecast for mac_2_8: 8.036614837387386 Avg. forecast for mac_4_16: 8.067234025374782 Avg. forecast for mac_8_32: 8.202621588370425 Avg. forecast for mac_16_64: 8.434827108432607 Avg. forecast for mac_32_128: 8.731612759969558 Avg. forecast for mac_64_256: 9.18820470763429 Avg. forecast for breakout_10: 9.273430193475184 Avg. forecast for breakout_20: 9.303238069489593 Avg. forecast for breakout_40: 9.239289775994772 Avg. forecast for breakout_80: 9.303273042345682 Avg. forecast for breakout_160: 9.197371607492794 Avg. forecast for breakout_320: 9.59132531301986 Avg. forecast for carry: 9.109571444401963 And a quick one for Rob, how come the system in Leveraged Trading is not using forecast diversification multiplier when using multiple rules? My combined forecast can't hit 10 either and I think that might be the issue as well, besides just individual ones coming short. It should be simple to run the backtest to see what the raw values for each rule are and come up with my own scaling factors, but I wanted to ask here first to see if that's something that should be recalculated regularly.[/CODE]
Actually, ignore this one. I went back to read again and realized that it’s accounted for in the suggested increase of the account level risk target. For those following, that’s table 51, page 212.
Great TTU episode! Thanks for airing this question, GAT. A quick follow-up: During your answer, you mentioned you allocate only 25% of your investable capital to TF. Did you mean the TF strategies in your futures system has a 25% of your capital allocated to them or does the 25% represent the capital allocation to the entire futures system (including the other strategies such carry, etc)? I ask because I am myself trying to figure out the best percentage allocation to different investments.