There is a ted talk about how intelligence is measured in AI and scientists have found that the number one thing is how many options the program leaves for itself. The more options, the more efficient it can navigate through problems. I would say this concept parallels to many aspects of life.
The following user says Thank You to Itchymoku for this post:
The only way you can have a trading plan is if you assume the market will act similar in a way going forward live as the market behaved previously - that is what a trading plan is. It's not making the assumption it always will, but that there is a higher probability at certain poiints (VAL, POC, VAH, order flow, whatever) it will act in a certain way and you are trying to capture that edge. The only way you can know if the market will react in a certain way is either with years of experience following the same plan, or automation of your thoughts. Either way, it is a repetition and not completely random, if there is no repetition then the market is completely chaotic and it won't be possible to make money from it.
Anyhow this deviates from the original question of the thread - you are of the anecdotal esoteric camp and I am from the evidence based camp, so we will never agree - so I do agree we agree to disagree.
I do believe in systems, so it is applicable - though your time frame is short term scalping of order flow and I would only apply systems to longer term capturing momentum from OTF trading, which is different - there are many hedge funds who use systems to do that.
I don't think mathematical is the only way, I believe in discretionary, infact, I think for short time frames discretionary is the only way to go except for arb as you mentioned, as order flow is hard to automate.
However, my logic is still sound - if you scale out twice or three times, and then charted your performance of each scale out and one particular scale out is clearly making you all the money, then you should not scale out of the other one since you DO have the skill to almost always pick an optimal. However, if the performance is intertwined, where one outperforms the other and then alternates, then you should scale out as that means you can't pick an optimal exit. Does that not make sense?
If I analyse my past performance and there is a clear persistent pattern, then I would try improve on it.
Also I think your definition of system is different to mine, it can have a defined entry point (order flow, breadth, etc) but not necessarily a defined exit point. Exit can be time, an opposing signal, etc but I am not talking specific targets. You are thinking data mined systems which I don't condone and won't work as you pointed out, I am thinking automation of a sound market hypothesis (order flow, market internals, value area, etc).
Last edited by PeakGrowth; December 24th, 2015 at 08:42 PM.
The following 4 users say Thank You to PeakGrowth for this post:
I'm afraid that indeed does not make sense. This actually reminds me of an old story that goes more or less like this: A hungry man was treated with a big dinner. After the 8th serving, he was finally done. He said to himself: had I known the 8th serving is the magic one, I probably should skip the first 7 servings and jump to the final serving directly.
In short term trading, it's hard to know which particular scale out is "making all the money". Without the other scale out, the "optimal" scale out will be our elusive 8th serving
Last edited by shzhning; December 24th, 2015 at 11:41 PM.
It's not hard to know, the issue is that you don't know because you don't understand it.
If you are all in and scale out, say 2 in and 1 out at POC and 1 out at VAH, you are effectively trading two separate methods at the same time, 1 with an exit at POC and 1 at VAH. If you have the data, you can easily see which exit type earned what money. If you are consistently losing money scaling out at POC, then once you only all out at VAH, you will increase your returns for the same amount of initial risk. If you can only earn consistent money by using both POC and VAH, then scaling out is the right method. This is called self improvement.
This works for all in scale out, however, if you decide to scale in and scale out, then it won't be reliable since subsequent scale in's will be dependent on the initial entry. You would then look at how much money your initial entry is making you, then your subsequent scale in is making you. If scale ins are consistently losing you money, the you should not scale in. Infact, if your initial entry isn't even making you money you should just go do something else.
This also applies if you don't use targets and you use price spikes. Perhaps you scale out at spike 1 and spike 2, you can then see if spike 1 or 2 is paying for half of your stop outs. If it's not, then the target is too close and you should only go all out at spike 2, then consider if spike 3 is worth chasing as the second scale out. This can all be quantified, if you know how to do it.
If you actively continue to use a particular scale out even though you know full well from your trading history you lose money from using it, then you are just giving money away.
@DionysusToast uses all in/scale out so what I said applies. However, I am not saying that his scaling out doesn't work or saying scaling out doesn't work in general, I am saying if your previous trading data tells you a certain scale out point is not profitable, it should not be used as your other scale outs would be just masking your sub optimal scaling - you can quite easily increase your profitability by smartly looking at your past trading, whether it be a backtest or live trading history.
If your answer to this is that, you don't know what rule you are using to exit and that you are just feeling the market on the way out, then you are just going by the seat of your pants or your years of experience, then in which case you keep doing your thing and none of this applies. Whether you are successful would be up to how good your flying pants are.
What gets measured can be improved, rather than deluding yourself that nothing can be improved because the market is unpredictable (which it is not) you would be better served to have an open mind.
The following user says Thank You to PeakGrowth for this post:
Is it 100% mechanical or 100% discretional or mixed. You are limiting yourself by saying that only 1 style works. If it does not work for you it is only you. Others are different. So are their systems and profits.
Datamining is a good tool for somebody. Past market action of course not very useful (or better to say: easy) for future actions but helps you a lot to learn execution (yourself). If "lucky" or not too hasty you may find something which can be turned into workable strategy or method. Someone finds hes problem by going through hes trades and calculating different kind of ratios & other mathematical stuff, other pal writes extensive trading log which will help him in this and that... numerous methods and tools available find the best ones which works for you. No matter if someone here whining it did not work for him so it will not work for you either which by the way is ridiculous thinking.
By the way one Poll option is missing: Scale in, all out (=what I am using)
It takes courage to be a Pig
Go With The Flow !
Last edited by Scalpguy; December 25th, 2015 at 04:52 AM.
The following user says Thank You to Scalpguy for this post:
peakgrowth I don't want to leave a bad taste because after all it's Christmas, this is merely debate. In all the years I've been on this forum and have talked to traders I've never really seen anyone do well long term with intraday automated systems based on back testing. @kevinkdog may be an exception.
Scaling in and out is almost like an insurance policy if you have the patience and diligence to not let greed and fear interfere.
Last edited by Itchymoku; December 25th, 2015 at 12:39 PM.
That's ok - you have already just proven my point with your own argument by saying kevindog is an exception; big mike is also an exception. Nobody said it was easy, there is a difference between possible and impossible and you just made my argument for me.
I might also point out in a professional environment if you come up with a strategy and you don't do proper back testing you will get laughed out the door. If you don't do back testing for yourself as a retail, just because nobody is laughing you out the door doesn't mean it's a good thing. If you just start firing off based on a theory that you just concocted, you have a low chance of achieving anything. You might have to come up with 30 theories before one of them works, do you have the time to manually forward test them each for years at a time, that is provided you don't sabotage yourself?
Once again, if you have read my posts and though about it, I have said back testing is used to prove it won't completely fail. A system that was back tested and failed completely has no chance of being a winning system, ever - therefore it is useful in determining if this is the case and whether you are wasting your time from the get go. Even a break even back test is useful, as you can use discretionary to turn it into a winning one.
Automation is simply ensuring your execution is 100% on point. Nobody (or I haven't) has said automation only works with stuff like RSI, and MA crosses or whatever. It can be applied to many types of tools discretionary traders use, such as cumulative delta, volume, tick, breadth, volume profile, delta momentum and so on and so forth. If your discretionary rules are specific enough, you can automate them provided you have the skill set to do so.
Once again, if you have read my posts and thought about it, I said scaling IS an insurance policy, as it helps your psychology and slows your burn rate if you are running a negative expectancy system. I also think it's a good idea if you are discretionary.
Feel free to debate me as much as you want, but I debate based on theory and evidence rather than anecdotes and analogies.
Anyhow, back on topic as I am done debating this topic in the inappropriate thread.
The simple truth is if your scale in/out points are not making you money (and you can find out by looking at your history), then you should not do it - you will instantly increase your expectancy by doing so. This is situational, everyone is different and should find out for themselves, hence, the answer to this thread is it depends.
This is not true at all. Prop traders get taught how to trade, not how to backtest. I know the management at a number of prop firms as well as many prop traders and they do not back test, they see trading as a skill that you develop over time, not as a 'system' you fine tune.
What they do is teach, review, journal, practice and (if you are lucky) mentor.
Your viewpoint is a systematic one - based on a belief set that the market is systematic. Not all share your view, nor can you really prove that it is an absolute as you seem to think.
I scale out and you say that I am definitely giving money back when I do that. That's not an assessment you are in a position to make.
But let me add some more to what I do....
- In a very slow market I will be all in/all out. Range trading
- With a more normal market, which on my market means runs that can last 30 mins to an hour and then reverse, I scale out because the target is unknown and depends purely on how other traders behave in terms of jumping on the move (as well as those that get stuck fading it).
- In a very fast market, I will scale in. Mostly because I know I can go offside quickly and that the market can put in a larger run, so I will de-risk the trade by starting small
Now - I could turn all this into statistics but that presumes that I can draw cause from the effect. I am a firm believer in logging but I log the state of the mark
You simply cannot automate certain styles of trading. There is too much nuance involved. It's not about rules, it's about the fact that automation will DUMB DOWN your trading style. You have to lose a lot of the nuance.
If you have any questions about the products or services provided, please send me a Private Message or use the futures.io "Ask Me Anything" thread
The following 5 users say Thank You to DionysusToast for this post: