System Development with acrary
- Thread starter acrary
- Start date
- Status
chewing on that question. my guru uses to say: one ounce of practise is worth tons of theory. build your own analysis tool i would recommend. everything that came from my side is easily done in excel.
i haven't seen the concept of stress corr elsewhere. on sharpe ratios and the portfolio effect: again your own spreadsheet will be the ounce ...
one of the things i found is that it is exponentially more exhausting to increase sharpe ratios the higher they are. moving from 1 to 1.5 is easier than to move from 1.5 to 2.25 though both is an increase of 50% in sharpe.
i haven't seen the concept of stress corr elsewhere. on sharpe ratios and the portfolio effect: again your own spreadsheet will be the ounce ...
one of the things i found is that it is exponentially more exhausting to increase sharpe ratios the higher they are. moving from 1 to 1.5 is easier than to move from 1.5 to 2.25 though both is an increase of 50% in sharpe.
Well, you pretty much summarized how to develop decent system on previous pages. To increase quality you can go academic route or take the practical use one. In either case some people will learn something.
If you feel like it you can post something on how to trade DAX.People keep asking me about it and I know didly ( about DAX )
Yes, + or - compared to today's close on major US cash indexes, e.g. SPX, DJIA.
The option is entered after today's settlement, and expires after tomorrow's settlement. The payoff at expiration is all or nothing, and you know exactly what the payoff will be when you purchase the option.
For "trend range days" or "key reversasl days" the price will be slightly biased, for example instead of purchase for $1 and payoff of $1, it may be 1.20:1.00. So you'll have to hit more than 50% if you're trying to pick "obviouis chart patterns". The flip side is for those "obvious chart patterns" you can model the market internals and predict when it will NOT follow the "obvious" direction. Lots of edge here as most traders just see "key reversal day" or some such... only problem is you may see only 1 of these situations per month.
The premium moves with after market, for example, if market moves a few points over night, in direction of the move risk:reward will be like 1.40:1.00 but if you want to take opposite the move you'll get the reverse.
You can also exit the option during the business day the MM makes money on the spread, so you're usually better off holding thorugh settlement. The only time I exit midday is if my model predicts a morning move and we're close to yesterday's close, meaning the market could basically close up or down. In that case, I know my prediction was wrong and I reduce risk by exiting with either small profit or loss.
That's one of my optimizations in my model, I don't just look at up or down, but if the move is predicted to occur morning or afternoon.
The only last issue is in order to have access to this, I have to commit a monthly minimum, so either few large trades or more frequent trades. My model makes a prediction every day and I make a trade about 60% of the time.
This option may be available to retail traders in 6-12 months. For now, to play the game requires low six figures and monthly minimum. I know I can nail this but need >60% and with consistency. With >70% the MM math changes from slow and steady to rocketship equity curve.
In summary: all you need to do is predict the direction of tomorrow's close. Magnitude does not matter. It can close 1 tick or 1000 ticks in your direction, and the payoff is the same. You know exactly what the costs are in advance when you purchase and you have ability to exit before expiration (which is next day's settlement), but may pay 10-15% in premium to do so.
Man,
I agree with you for all cases except for what I described. You didn't read why hit ratio is of utmost important to the position I described.
Hit ratio does not matter if magnitude of a trade is part of the result (whether you made 10 ticks on a trade or 1000 ticks). Thus trend followers can be wrong more often but their wins are larger, hence a variety of other ratios are used.
The position I described has almost a 1:1 correlation between hit ratio and payoff ratio, the difference being cost and some premium if you enter a trade late.
Imagine you are flipping a biased coin where you win 60% of the time. The hit ratio is the same as the payoff ratio for all intenstive purposes. The only difference is bet sizing.
This is why the trading vehicle I described is a whole new way of trading using neural networks, which was the part of acary's post I was replying too (that with NN's and/or other methods he can predict direction >70% of the time).
Acary uses direction prediction as the FIRST STEP in trading strategy. What I'm saying is I have a way to use direction prediction (what you call "hit ratio in this case") as the FIRST AND FINAL step. No profit targets, no stop loss.
Imagine you have a neighbor who has unlimited capital and is willing to quote you price every day on whether tomorrow's close will be higher or lower. Doesn't matter how far the market travels or how often it oscilates over whatever stop loss a daytrader would've had in place. It's a fixed wager on up or down close.
It's a whole new way to trade the game when you remove magnitude and intraday movement out of equation.
Again, I am saying I personally cannot acheive >70% for any long period of time (e.g. 100 days) using a variety of advanced NN and other methods. I have never seen anyone who has. I've seen some potential models that ran well for ~30 days then a phase shift occurs in the market. I can acheive 60%. I've proposed a no-risk deals to several pro's who claim they can pick prediction right >70% but it never panned out. The difference between 60% and 70% is two orders of magnitude for me.
The reason it matters is I don't think this degree of inefficiency will exist for more than another 24-months once these type of exotics become available to retail traders. If you are familiar with sports betting, imagine Super Bowl props being made available for every game very day of the year.
dottom
sorry for not properly addressing your post. first i thought you used the option issue just as an analogy. now i see that you meant it iterally, meaning you actually buy a derivative. i see your point now.
62% hit ratio and 1.0 payoff makes a profit factor of 1.63 and if you bet about 150 times a year that makes sense. i wonder how your provider hedges these options in extreme scenarios, when the market starts moving heavily - well it is major indices. should be liquid all the time near the money.
what is your concern on the retail accessbility? retail money won't be too smart, so your prediction edge won't fade due to that. even more this money should reduce the cost of the options. actually i wonder more about the people offering the options. sorry if i appear dumb (well i truly am dumb when it comes to hedging exotics
), but someone has to swallow the big outlier events. i would think it is the hedger (constantly adjusting hedges in a heavily trending market costs a lot i imagine) and i would think these outlier cost should be reflected in the option price/spread/commission whatever.
my point is that within a perfect market these options would be priced in a way that completely offset the risk. we all know that you can easily have trading account with highest possible hit ratios and you still go bust ... selling options. 80something hitratio, but payoff in the ground. resulting in net loss. if you can predict above random and someone grants you payoff of 1, there is some flaw in that ... and i would think it is that they can offer this is in a low vola environment and will be set off once we see a spike and their hedges take them on the street.
in 98' a trading desk offered options on hedge fund pools. they used hedge fund historic vola plus some buffer. this kind of product vanished by summer. we actually were quite lucky to have a very cheaply bought option on our book for years.
sorry for not properly addressing your post. first i thought you used the option issue just as an analogy. now i see that you meant it iterally, meaning you actually buy a derivative. i see your point now.
62% hit ratio and 1.0 payoff makes a profit factor of 1.63 and if you bet about 150 times a year that makes sense. i wonder how your provider hedges these options in extreme scenarios, when the market starts moving heavily - well it is major indices. should be liquid all the time near the money.
what is your concern on the retail accessbility? retail money won't be too smart, so your prediction edge won't fade due to that. even more this money should reduce the cost of the options. actually i wonder more about the people offering the options. sorry if i appear dumb (well i truly am dumb when it comes to hedging exotics
), but someone has to swallow the big outlier events. i would think it is the hedger (constantly adjusting hedges in a heavily trending market costs a lot i imagine) and i would think these outlier cost should be reflected in the option price/spread/commission whatever. my point is that within a perfect market these options would be priced in a way that completely offset the risk. we all know that you can easily have trading account with highest possible hit ratios and you still go bust ... selling options. 80something hitratio, but payoff in the ground. resulting in net loss. if you can predict above random and someone grants you payoff of 1, there is some flaw in that ... and i would think it is that they can offer this is in a low vola environment and will be set off once we see a spike and their hedges take them on the street.
in 98' a trading desk offered options on hedge fund pools. they used hedge fund historic vola plus some buffer. this kind of product vanished by summer. we actually were quite lucky to have a very cheaply bought option on our book for years.
1/ Do you have to be right on all major indexes or it is possible to chose and then use just one ? NDX seems to be the best suited for your deal. If only NDX or QQQQ can be used it sound very possible to go over 60%.
2/ I think that there is more to it( in your method ) then just being at or close to to yesterday close to exit intraday.
If being at or close to yesterday close is your only reason for exiting, this can be easily improved. I am sure that you saw many intraday reversals at price levels of yesterday close and those can be predicted.
my e mail is walter904us@yahoo.com if you want to pursue this further.
Now, let's go fishing !
- Status