First time posting here and I'm not sure if this is the right category; apologies if it is. I've been interested in trading (primarily Forex) for a number of years and I just became serious about making it a career. I've been backtesting a system that I think is pretty good, but I would like some other opinions on the results. Below are some of the results; any input is very much appreciated!
2005-2017
- Annualized Return: 46.26%
- Win Rate: 55.54%
- Number of Trades: 2,618
- Max Drawdown: -28.68%
- Avg. Max Annual Drawdown: -18.26%
- CALMAR Ratio: 1.61
- Sterling Ratio: 1.64
- Sharpe Ratio: 1.73
2015-2017 (last 3 years)
- Annualized Return: 60.46%
- Win Rate: 56.36%
- Number of Trades: 621
- Max Drawdown: -26.46%
- Avg. Annual Drawdown: -19.78%
- CALMAR Ratio: 2.28
- Sterling Ratio: 2.03
- Sharpe Ratio: 2.60
Hate to admit, back testing doesn't mean anything, to either a hedge fund or a prop firm and you didn't really clarify what your strategy is... Just numbers... Based on your annualized returns, you're pretty much stating that you're beating out all these PhDs quants at hedge funds... Sorry but unless this is real P/L risk, it's really hard to give you an answer and I wouldn't quit your day job...
Thanks for the insight :)
If these numbers hold up, at least somewhat similarly, in real world trading, how would these metrics stack up, particularly the ratios?
put your money where your strategy is and trade it...and after a year see where you are at.
also, i'd suggest you go code this in python in quantopian....and see if it still plays as you hope.
What's your PnL per trade? And while we are at it, what sort of fill assumptions did you use? :D
##########also, how curve-fit is your system?
Not as big a problem with forex but strategies don't typically scale cleanly.
Backtesting is also...questionable most of the time. See LTCM for an example.
Not trying to sound harsh, but like what the earlier poster said, a strong backtest result alone doesn’t mean anything.
It’s a start but give me no constraints and I’ll give you a 2.0 Sharpe backtest result in 1 hour.
It takes time but here are some things to consider.
Think those through and see what you get.
Thank you everyone for your responses! I will get to work on everyone's suggestions.
But, my original question, how do the results of my backtest compare against top traders, hedge funds, etc?
the BEST traders on the planet avg 25% annualized return over a period of many years.
(yes, some traders make 300% or some crazy high number a few times...but over a large sample size of years those numbers come down...especially as AUM increases into the billions and liquidity becomes an issue)
Now, what are the odds that YOU stumbles upon a better strategy? 0.001%
The best average a lot more than 25% per year. Ten years ago 200-400% was possible. Tougher now. Soros was doing 30-50% a year at huge scale.
Definitely a lot tougher today....but all the big shops use to do far higher (the one shops that converted to a family office put up 50% on a billion).
Thanks!
What about some of the other ratios, drawdowns, etc.?
the larger multi-manager funds typically have a max drawdown of 10%...meaning if you lose 10%...you are fired. Some split that in half...lose 5% and your capital is cut in half...lose another 5%...and you are fired. Some have those numbers as 4% and 8%. Most also have 1% max allowable intraday drawdown.
So, you would most likely need to significantly reduce your size/leverage by 3x if you ever wanted to take such a strategy to a hedge fund where you are a portfolio manager.
This would drop your returns to 15%-20%...still good....not stratospheric like 46% or 60%. I'm curious how that would affect your sharpe...probably drops closer to 1 from your volatility.
Check out the forum Nuclear Phynance btw. Incredibly smart quants on there and a ton of info that goes much further than what youll find here. It's literally all super nerdy PHDs. Just bounce through that forum and you'll see what we mean re: issues with backtesting.
If it were to hold out of sample and can trade size then yeah the stats are decent and likely would be traded in some way at a lot of places (whether as a dependent variable going into some larger system or solo). But I'm going to guess you haven't modeled the market impact and might not scale easily. I have a signal that's 4 sharpe/60% levered returns that we don't bother trading at work because it won't scale and can't net with other stuff so not worth the hassle.
What would you consider trading "size" to be? How big does something need to be able to scale to that it is considered worth the time and effort?
no hard and fast rules but some ballparks.. 1 M in pnl with hope can grow beyond that if it's going takes some tech work and doesn't net with current stuff.
if hiring someone that's bringing it along and they want to be directly tied to pnl then 5 M pnl and uncorrelated to existing book
if something I came up with and can net with existing stuff then as long as it lowers turn or raises sharpe then it'll be thrown into the mix (or I'll just throw it in to avoid cherry picking/overfitting), so for example not trading the 4 sharpe low capacity thing but do have 1 sharpe low turn/uncorrelated signals that are traded in the book
currently running a book at a multi-manager hedge fund with payout based on sharpe so ultimately look at impact on payout
net dollar income on one axis and sharpe based on ROC (net income / allocated GMV) on the other
wait...so if you make 30mm profit on a 200mm capital allocation....as an example, if your sharpe is 1, you could get 10%, and if your sharpe is 2 you could get 15%? (something along those lines)....even tho in both cases your $$ pnl is the same....and in both cases you have stayed within the firms risk mgmt rules?
yep exactly.
fundamentally the reason being the fund can use more leverage if a higher sharpe is achieved and so incentive for it
Are sharpes really something you can leverage for nowadays? With all the algos in trading that seems like a great way to get blow out.
pretty much have to in order to get decent returns, most large equity market neutral quant funds make like 3-6% a year on GMV but levered 10x makes for 30-60% on capital
Seems rather dangerous in today’s market. Has a tendency to break in unexpected ways.
A highly leveraged oil arb fund blew up a few months ago. And of course the Vix earlier in the year. Some of the moves were completely non-fundamental.
very dangerous if you don't know what you're doing, been doing this for 15 years... when the vix spiked earlier this year we were printing money and 2008 was our best year.
we hedge out most fundamental factors
can you elaborate on this? / if you are able to provide some literature / books that cover this in detail?
find alpha (the hard part), run optimization to neutralize/constrain risk factors while taking into account market impact/tcost/etc so one hopefully is putting on trades that actually make money, and then execute
the shorter termthe alpha the less the risk factors matter but the harder to execute (especially with the likes of virtu who probably have more invested in tech infrastructure than a lot funds have in capital) ' the longer term the alpha (especially overnight to days) then controlling risk becomes incredibly important since if you don't gonna end up trading things one doesn't intend to
so overall rather simple but devil is in the details and there isn't any literature that goes into sufficient detail to do it from scratch and most stuff out there is borderline useless
but a few books that are mildly helpful: Active Portfolio Management by Grinold/Kahn, Quantitative Equity Portfolio Management by Chincarini/Kim, Advances in Financial Machine Learning by Prado
.
Watch the drawdown/leverage. Also, “there is no such thing as a bad backtest”, so be mindful that these metrics can be misleading. As somebody posted, many firm try to stay within the 10% mark and under for drawdowns, and returns in the teens + are actually typically very scalable if your risk is in line. I strongly recommend never solely relying on one strategy. Just like a portfolio, you should have a basket of non correlated return streams / strategies. Lastly, put money behind your ideas, and automate what you can. Pick up some programming if you haven’t done so already.
Some people commented that funds have a 10% max drawdown before you're fired. Do they also typically have restrictions on how much you can risk per trade?
Et beatae expedita est sed ea rem minus. Error nisi quisquam distinctio itaque modi.
Inventore ut tempore non consectetur. Maxime dolore quia nostrum eveniet. Enim nam nisi architecto ut quo. Ratione qui dolorem et consequatur dolorem sit. Facere qui culpa cupiditate occaecati dolor eos quaerat.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...
Culpa architecto hic et dolorem exercitationem dolorem ipsa quisquam. Tempora at quis rerum perferendis voluptatum. Nihil voluptatem voluptate vitae laudantium quaerat.
Tempore qui perspiciatis reprehenderit. Dicta doloremque consequatur eaque nemo. Qui inventore quis est sed fugit. Ut accusamus velit suscipit nesciunt impedit.
Cupiditate officia quis reprehenderit. Doloremque odit aliquid libero sunt quas sapiente.
Qui earum consectetur quos. Eaque et repellendus delectus aut earum earum ipsam. Omnis labore nesciunt ut commodi tenetur.