Trade another day through Systematic Trading | Page 3 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Another Day, Another Chair, Another Day in the Office 203 replies

Systematic trading 1,576 replies

Nonparametric Association Measures in Systematic Trading 57 replies

A Systematic Approach to Markets 68 replies

Systematic Trading vs. Discretionary Trading 14 replies

Attachments: Trade another day through Systematic Trading
Exit Attachments

Trade another day through Systematic Trading

  • #41
  • Edited 8:43pm May 1, 2019 7:41pm | Edited 8:43pm
From my experience, "linear" systems always have to be reviewed and tweaked, because soon or later they are doomed.

That's what all these quantitative hedge funds are doing with their black boxes.
And they have some really complex systems - to come even close to make some money.

It isn't easy at all.
The competition doesn't sleep.

Edit: ya, back to topic.
  • #42
  • Edited 7:35am May 2, 2019 12:12am | Edited 7:35am
Quoting The-Flipper
From my experience, "linear" systems always have to be reviewed and tweaked, because soon or later they are doomed. That's what all these quantitative hedge funds are doing with their black boxes. And they have some really complex systems - to come even close to make some money. It isn't easy at all. The competition doesn't sleep. Edit: ya, back to topic.
I agree with you.

A few classical linear systems do work as long as the money management part is coded well.

It is quite tough translating discretionary ways of money management which includes positions, lots, pyramiding, averaging, reining in drawdown, maintenance margin levels % , trade entry and exit accuracy, etc.

Because I don't have programming or even Excel skills, I love to trade non-linear instruments like options. My brain is my EA and paper and pencil are my tools.
Besides I don't need as deep pockets as linear instruments like futures and OTC fx need.

People are unrealistically seeking fortunes without having enough capital and with a mindset that is not compatible with facing random trading. System or no sytem they are bound to fail if they don't fufiill the basics.
Practice makes a person perfect
  • #43
  • May 2, 2019 4:19am May 2, 2019 4:19am
Quoting alphadude
{quote} hi V, I can't recall exactly how many strategies I retire; but quite a bit. On average; perhaps one strategy is dead every 3 months; or so. Most dead strategies was traded over a year or more. Some are not dead 100%; but need improving. Think of it as Theasurous ship; over time; they replaced every single spare part in the ship. Is the ship still the same ship; or completely different. Last year; I have shared my journey over an 11 months period; reaching 100% return on the Verbal Dierreia challenge. There are few tidbits there if you revisit...
Hi alpha,

Do you evaluate the strategies in a monthly basis? What parameters do you check in order to dismiss a strat?

David
  • #44
  • May 2, 2019 6:35am May 2, 2019 6:35am
Quoting DavidRP
{quote} Hi alpha, Do you evaluate the strategies in a monthly basis? What parameters do you check in order to dismiss a strat? David
any of the following:
- negative return in the first month of trading
- drawdown reached 2x strategy drawdown in its backtest
- daily live trades does not match backtest trades. this is evaluated on daily basis at EOD
  • #45
  • Edited 2:59pm May 2, 2019 2:38pm | Edited 2:59pm
Quoting alphadude
{quote} any of the following: - negative return in the first month of trading - drawdown reached 2x strategy drawdown in its backtest - daily live trades does not match backtest trades. this is evaluated on daily basis at EOD
I dont know how I feel about this man. You guys are more knowledgeable with running algos in live than I am but I am going in with the expectation that any algo has to work for atleast 6 months. If I have to analyze performance with backtests on a daily basis, I am just better off trading manually. Is that what others are also experiencing Nik? is this the same in your case? Are you evaluating order flow between live and backtest daily/weekly? That's like WORKING TWO JOBS lol I want to enjoy the money my algo makes and not the other way around :-(

EDIT: Perhaps, I should clarify. I am not setting the expectation that markets don't change and algo should work as is without intervention. I think if the strategy is based on individual trade RR then it forces you to reevaluate more frequently and hence my reason to stay away from entry enforced RR systems. I like the RR based on equity stops and profit taking. it is much better model to assess strategy outcome imo... along with some basic stats. And also, if my understanding of "Market is universal and neutral" is correct, any strategy designed to confirm to the universality principle should work/survive a little bit longer... i would hope :-
Staying in my lane...
  • #46
  • May 2, 2019 2:57pm May 2, 2019 2:57pm
Quoting VEEFX
{quote} I dont know how I feel about this man. You guys are more knowledgeable with running algos in live than I am but I am going in with the expectation that any algo has to work for atleast 6 months. If I have to analyze performance with backtests on a daily basis, I am just better off trading manually. Is that what others are also experiencing Nik? is this the same in your case? Are you evaluating order flow between live and backtest daily/weekly? That's like WORKING TWO JOBS lol I want to enjoy the money my algo makes and not the other way...
lol. its hard work man.

the idea here is to test that the program you have written is doing what it supposed to do.

at the end of day, i download tick data, and run the backtest on it. the trades should match live trades. if not then you got a bug somewhere.

in fact i do this multiple times a day.
  • #47
  • May 2, 2019 3:01pm May 2, 2019 3:01pm
Quoting alphadude
{quote} lol. its hard work man. the idea here is to test that the program you have written is doing what it supposed to do. at the end of day, i download tick data, and run the backtest on it. the trades should match live trades. if not then you got a bug somewhere. in fact i do this multiple times a day.
I just edited my post... that is different from ensuring if the algo is performing as per the backtesting results. Bugs are a number one killer so agree, code has to be bug free and testing it frequently with live data is all good prudence to me.
Staying in my lane...
  • #48
  • May 2, 2019 3:04pm May 2, 2019 3:04pm
Quoting VEEFX
{quote} I just edited my post... that is different from ensuring if the algo is performing as per the backtesting results. Bugs are a number one killer so agree, code has to be bug free and testing it frequently with live data is all good prudence to me.
and just to say it.... on most days, I am trading manually in sim and then run my algo in sim and then again my ago in tick mode for few hours... doing this shit for over a month. this is HARD WORK.. there is no doubt about it. BUT, I am/was hoping this HARD WORK is mostly upfront.... and not recursive on a daily/weekly basis. !
Staying in my lane...
  • #49
  • May 2, 2019 3:18pm May 2, 2019 3:18pm
Quoting VEEFX
{quote} and just to say it.... on most days, I am trading manually in sim and then run my algo in sim and then again my ago in tick mode for few hours... doing this shit for over a month. this is HARD WORK.. there is no doubt about it. BUT, I am/was hoping this HARD WORK is mostly upfront.... and not recursive on a daily/weekly basis. !
you could have a bug anywhere; including in the market data itself.

I once had a problem on the VPS. the tick data was received by the VPS with huge latency (ranging from 1second to 10 seconds); due to DOS attacks (denial of service attacks on the VPS).

since then I started using dedicated server; instead of VPS; which is being shared by many; and becomes a target of hackers.

your program has to validate every single tick data when it is trading live.
  • #50
  • May 2, 2019 3:25pm May 2, 2019 3:25pm
Quoting alphadude
{quote} you could have a bug anywhere; including in the market data itself. I once had a problem on the VPS. the tick data was received by the VPS with huge latency (ranging from 1second to 10 seconds); due to DOS attacks (denial of service attacks on the VPS). since then I started using dedicated server; instead of VPS; which is being shared by many; and becomes a target of hackers. your program has to validate every single tick data when it is trading live.
maybe in your case this is true as you have built a custom solution in Java to receive tick data. makes sense. In other instances like Sierrra where data feed connectivity and validation controls are already part of the software package, we don't need to worry about it in our algo correct?

And how do I even validate tick data in real time in live trading? you lost me there!
Staying in my lane...
  • #51
  • May 2, 2019 3:28pm May 2, 2019 3:28pm
Quoting VEEFX
{quote} maybe in your case this is true as you have built a custom solution in Java to receive tick data. makes sense. In other instances like Sierrra where data feed connectivity and validation controls are already part of the software package, we don't need to worry about it in our algo correct? And how do I even validate tick data in real time in live trading? you lost me there!
and just for the record... anyone running their algo on a public VPS is dumb as a rock!... especially the ones offered by your "broker". Unless you are taking advantage of colo services to address latency, a simple server in your garage will do (for me and hire a dude from India to remotely monitor my algo)
Staying in my lane...
  • #52
  • May 2, 2019 3:28pm May 2, 2019 3:28pm
Quoting VEEFX
{quote} maybe in your case this is true as you have built a custom solution in Java to receive tick data. makes sense. In other instances like Sierrra where data feed connectivity and validation controls are already part of the software package, we don't need to worry about it in our algo correct? And how do I even validate tick data in real time in live trading? you lost me there!
probably correct. but I don't think latency is accounted for by sierra or any other software like MT4 for example.

when I used to trade in MT4; I have an EA that persist tickdata of the instruments I trade to a flat file; which I use later for backtesting at the end of the day. you can do the same thing in Sierra or any other software.

however; don't worry about it too much at the beginning of your jounrny. Once you have a profitable algo generating $$$ while you are asleep; start improving it by looking at all other aspects.

Slippage and brokerage are another monsters that you need to reduce down the road.
  • #53
  • May 2, 2019 3:37pm May 2, 2019 3:37pm
Quoting alphadude
{quote} probably correct. but I don't think latency is accounted for by sierra or any other software like MT4 for example. when I used to trade in MT4; I have an EA that persist tickdata of the instruments I trade to a flat file; which I use later for backtesting at the end of the day. you can do the same thing in Sierra or any other software.
yes..Sierra has all those controls and more when it comes to robustness to data service provider connectivity. it also supports persistence of tick data into a data file for offline validation etc. If MT4 was ONLY available for download from their servers instead of have brokers customized their software for download, I would have no problem with MT4. Similarly, I would not use Sierra if it is downloadable from AMP or other brokers. That falls under a "lack of trust factor" for me. the software trading platform MUST be independent from the clearing firms who do the order matching/routing/executions. There needs to be transparency i.e. open source if you will to catch scams and inconsistencies
Staying in my lane...
  • #54
  • May 2, 2019 3:40pm May 2, 2019 3:40pm
Quoting alphadude
{quote} probably correct. but I don't think latency is accounted for by sierra or any other software like MT4 for example. when I used to trade in MT4; I have an EA that persist tickdata of the instruments I trade to a flat file; which I use later for backtesting at the end of the day. you can do the same thing in Sierra or any other software. however; don't worry about it too much at the beginning of your jounrny. Once you have a profitable algo generating $$$ while you are asleep; start improving it by looking at all other aspects. Slippage and...
Slippage and brokerage are another monsters -they don't bother me because to me, an algo is a failed algo if it does not pass with 2, 3 or 5 times the trading costs involved. for NQ, that means atleast $10-$15 per contract per trade to substract from trade pnl... that's how I put rigor and that is the main reason why I have yet to find EDGE in NQ.
Staying in my lane...
  • #55
  • May 2, 2019 4:31pm May 2, 2019 4:31pm
Quoting VEEFX
{quote} Slippage and brokerage are another monsters -they don't bother me because to me, an algo is a failed algo if it does not pass with 2, 3 or 5 times the trading costs involved. for NQ, that means atleast $10-$15 per contract per trade to substract from trade pnl... that's how I put rigor and that is the main reason why I have yet to find EDGE in NQ.
that i am afraid i disagree with.

transaction cost is a profit killer. as a business, you have to account for everything. be chinese if that is required to make money for your cruise
  • #56
  • May 2, 2019 4:51pm May 2, 2019 4:51pm
Quoting alphadude
{quote} that i am afraid i disagree with. transaction cost is a profit killer. as a business, you have to account for everything. be chinese if that is required to make money for your cruise
ofc... we all got to do everything to reduce costs. I am referring to factoring trading costs in to the equation in edge finding mission i.e. I won't rule out a strategy with say 55% winrate on 1RR risk model if it nicely pays me more than x times the trading costs involved. i.e. TradingCostMultiple as a parameter into my strategy that I can tweak and put the algo thru the ringer to assess the various possibilities.
Staying in my lane...
  • #57
  • May 2, 2019 9:59pm May 2, 2019 9:59pm
What was the longest period your bot was running successfully without touching or tweaking it?

Anyone?
  • #58
  • May 3, 2019 11:28pm May 3, 2019 11:28pm
Anyone here trades stat arb models instead of indicator based systems?

I have developed one (out of curiosity that involves basically the DOLLAR (Index)) and it works extremely well.
  • #59
  • May 4, 2019 12:08am May 4, 2019 12:08am
Quoting The-Flipper
Anyone here trades stat arb models instead of indicator based systems? I have developed one (out of curiosity that involves basically the DOLLAR (Index)) and it works extremely well.
If all price action crap fails for my algo on Nasdaq index, this is my next model to try (but I will be doing the modeling in real time and not based on some historical statistical anomalies or edge). I went the same route in spot fx correlations but all in real time based on strength modeling across 28 pairs (but also using simple price action candlestick shit because individual trade outcome did not matter in my case) but never designed for full automation. I was compiling mql atleast 3-4 times a day changing hardcoded parameters as I had over 350+ (some were designed to confuse MMs to prevent manipulation). If stat arg fails then I will work on a real money flow model based on global currency flows using UJ, dow/es/nq, gold, oil and tbills.... as a research project to keep me sharp:-)

I seriously doubt this site is the place for advanced quant topics. I asked a question earlier if anyone knows how to calculate a value of 1Tick of NQ in regards to market capitalization changes in nasdaq 100 index and I got no response whatsoever lol

Is your model for spotfx pairs?
Staying in my lane...
  • #60
  • May 4, 2019 12:51am May 4, 2019 12:51am
Quoting VEEFX
{quote}I asked a question earlier if anyone knows how to calculate a value of 1Tick of NQ in regards to market capitalization changes in nasdaq 100 index and I got no response whatsoever lol Is your model for spotfx pairs?
Is there no software that does it?
Playing with numbers can't be that hard.

Yes, it's for ALL major FX pairs and one derivative designed especially for the EURAUD and EURNZD
0 traders viewing now