Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
@bobbakerr, that's not very funny, but that's interesting to know, for people who don't already knows it: back and forward testing, simulation with a real-time data feed, can be very different from real life results...
The VPS I gave you to test is in Chicago, not in NY, so you could be 15-18ms faster with a server in NY.
But it will change your results (or not a lot): 140 R.T. in 97 minutes, this is not HFT, but it's impossible for an individual trader who pays usual commissions and use this hardware/software/network combo to make money. Losing "only" $1200 is a good result for 140 R.T. !
Individual auto traders can't play the same game as the professionals: they are working with micro-seconds latency, and we are dealing with milli-seconds (X 1000), our commissions are much much bigger, ... For an individual, a good bot will take few trades per hour, sometimes few per session, but not hundreds or thousands, it's impossible to make money with too many trades, the commissions and the slippage will destroy your account very quickly.
Thanks for your very informative response Bob, really appreciate it, and I already learned quite a lot from it.
That’s indeed brutal, and very hard to overcome. Personally I find it very disturbing to experience so much slippage – I thought that slippage in real-time would be a lot less, especially since you use the Rithmic data feed, MultiCharts and co-location. Did you try the system on other markets or was it designed specifically for CL?
Btw, thanks for the further information of the system performance and the big difference between real-time and backtested. Quite insightful and very important.
The theoretical results are quite impressive, and in real-time trading with your 2 cents Limit Order in effect you already give up a whole bar of a potential move. Is that 2 cent move in the beginning the ‘make or break’ component of your strategy, or do you think not getting filled and other time consuming activities (like hardware or software) are to blame? (I tried to make a diagram, see attachment, with the time leaks but I don't understand where all those seconds are coming from. I'd love to hear your opinion on this. ).
I also noticed that the volume in CL on 2 cent range bars is somewhat small – some bars don’t even have 50 contracts traded during RTH. Off course, for two seconds that’s quite a lot, but perhaps the liquidity is not enough if you are retail trader who wants to squeeze in his order between the other traders?
Wow, that’s not good. Did you uncover the source of these discrepancies? Have you contacted MC support, because those missing trades, especially with an active strategy as yours, are perhaps ‘killing’ the whole strategy.
Thanks Bob, very insightful. Thanks. If you had to make an new strategy, let’s say on the CL, what time frame would you use?
Thanks for your interesting comment Sam. Can you perhaps comment on the latency of the VPS that you provided for Bob?
This because Bobs mentions some trades weren't filled for 5 seconds or more, so I'm wondering where all those seconds are going. I’m quite a beginner at this, and I’ve attached a basic theoretical model from a book about high frequency trading to better understand why it takes “so long” to get filled.
If I may assume that the VPS has a latency of 100 ms, and the broker of Bob is DMA, which let’s say takes another 100 ms to send the order, and MultiCharts with the Rithmic datafeed takes 2 seconds to process the strategy and submit the order (again, just assuming). Would the other 2.8 seconds than just be the time it takes to fill the order which is at the exchange? Or am I missing a crucial bottleneck with these steps?
I'm very curious what both of you think about this, especially since the (relative) slow fills are breaking this strategy.
It's hard to find what was the bottleneck, if we have bottleneck here, without all details and logs (signal at xx:xx:xx.xxx, order sent at yy:yy:yy.yyy, received at zz:zz:zz.zzz, blah blah blah). The only think I can say is:
- from the VPS to NY we have 17 ms to 20 ms latency (vs 1 ms to2 ms to CME),
- the cpu load of @bobbakerr was a bit higher than the other VPS on this server (25% of two Xeon cores), but that's not too heavy, and the overall loadof the physical server was low (10% for 8 cores)
- his VPS did not use all the physical memory available (a bit more than 1GB for 1.5GB available)
- the disk i/o and network i/o were also quite low,
so no bottleneck on the VPS side...
there is not enough information on yours posts to determine if the -$1200 was because of the strategy produced bad signals, beacuse of slippage (and how many ticks), if because of the platform and execution network(TS is slow regardless of how close you get), or if because of transactional costs... as an example, I am still testing eminishark trader and depending on the type of timeframe or bar I use for signal generation it will loose money because of the commissions... in your case, it might be the same, but hard to tell without clear data that we can look at to direct you better... you would need to provide trade statistics.
As to the actual lost, as long as you are properly capitalized it is not that big of a deal IMO.. so though it was silly to test with real cash immediately, I have done the same so I can understand... I would recommend you look at IBKR given their simm/demo account is actually better than most things out there... you can setup MC/NT/etc to trade against it using any other live feed to generate the live signals that you so desire.
No, I have only tried it on CL. Perhaps a less volatile market with more volume would work. I'll give ES a try.
I think the fact that I am merely a retail trader ('little guy') is what hurt me. I don't have the access or 'power' of the professionals. I need to either try for bigger moves (with bigger losses also) or like I said above, something less volatile and with more volume.
No. The last time I contacted MC with a question about this, it took two tries to get a response. Then the question was not directly answered. So I'm not going to contact them about this again.
I have no idea yet. Maybe around 5 minutes? Or around 12- or 13-cent Range Bars?
off the bat... your edge is too tiny and will always be taken away by slippage... also, your broker is too expensive.. but a lower cost one or even with prof fees that wont make a difference to you.. not to mention that you still have to add into your transaction costs whatever infrastructure costs you have to cover before you make some $$$... lets call it the carry cost...
anyhow..
you are paying 11% more comms than the lowest cost broker we discuss on the forum, Crossland, you should address your comms with your current broker..
with prof fees (assuming $1.25RT/car) you would have lost less, but still loss given your loss per contract was around $4.04, which tells me that your edge is less than 50% winners to certain extent.. again, not enough data being provided by you for any of us to give you some constructive feedback, so I am taking stabs at the dark with my comments..
if you used NT7 for this trades, it would help if you publish the report for the day.. that does not show the strategy so no risk to you there.. if you used TS8/9, then it has the same thing .. export the report and remove your account data... if MC, same thing, export, remove... showing the trade performance does not mean you expose your strategy logic..