Pages:
Author

Topic: Automated Trading Contest (Prize 0.5 BTC) / Best automated strategies - page 3. (Read 50214 times)

legendary
Activity: 1974
Merit: 1077
^ Will code for Bitcoins
Here is the code I also posted on the cryptotrader forum:

https://cryptotrader.org/backtests/3uJPMJdhY9Ct4S6i8

Results: >2million for the year depending on current exchange rate (the linked backtest is only up until 21 Dec - adjust as appropriate)

It combines a rewritten Ichimoku algorithm with Heikin-Ashi candlesticks and Parabolic SAR. I share it here because it is a combination of my own work and that of a few others.

Please review and recommend further improvements!

Nice Ichimoku re-write, interesting.

How much is gained with those Heikin-Ashi candles compared to regular ones, have you intensively back-tested that?
newbie
Activity: 7
Merit: 0
Here is the code I also posted on the cryptotrader forum:

https://cryptotrader.org/backtests/3uJPMJdhY9Ct4S6i8

Results: >2million for the year depending on current exchange rate (the linked backtest is only up until 21 Dec - adjust as appropriate)

It combines a rewritten Ichimoku algorithm with Heikin-Ashi candlesticks and Parabolic SAR. I share it here because it is a combination of my own work and that of a few others.

Please review and recommend further improvements!
newbie
Activity: 15
Merit: 0
I've been playing around with my own version of the Ichimoku algorithm (inspired by medji's) and am finding that there is certainly no one-size-fits-all setting for it. Settings that have been tuned to work great over the full year turn out to be not so great when run over a more recent period such from 1st December.

In hindsight this is pretty obvious, but it highlights the fact you need to continually tweak your algorithm for current market conditions to get best results.

Anyway, I'm too late to enter but my current best for period 1st January to 21st December on MtGox with 0.55% fee is:

$1757649.52 (2630.01512 BTC)

Impressive Smiley I haven't played much with the full year performance since my k1b.. which is shy of 1.67M...

Thanks. I'm not actually trading with that one as its performance over more recent periods isn't as good as other variants I have. Currently I have two bots trading live against MtGox and BTC-e (each tuned to the particular exchange) with around 1 BTC each. I would like to throw some more BTC at them each, but want to see how they get on for a month or so first... it's still only early days so no reports on actual results yet...

Same here... and right now I am working on getting an offline backtest system.
newbie
Activity: 7
Merit: 0
I've been playing around with my own version of the Ichimoku algorithm (inspired by medji's) and am finding that there is certainly no one-size-fits-all setting for it. Settings that have been tuned to work great over the full year turn out to be not so great when run over a more recent period such from 1st December.

In hindsight this is pretty obvious, but it highlights the fact you need to continually tweak your algorithm for current market conditions to get best results.

Anyway, I'm too late to enter but my current best for period 1st January to 21st December on MtGox with 0.55% fee is:

$1757649.52 (2630.01512 BTC)

Impressive Smiley I haven't played much with the full year performance since my k1b.. which is shy of 1.67M...

Thanks. I'm not actually trading with that one as its performance over more recent periods isn't as good as other variants I have. Currently I have two bots trading live against MtGox and BTC-e (each tuned to the particular exchange) with around 1 BTC each. I would like to throw some more BTC at them each, but want to see how they get on for a month or so first... it's still only early days so no reports on actual results yet...
newbie
Activity: 15
Merit: 0
I've been playing around with my own version of the Ichimoku algorithm (inspired by medji's) and am finding that there is certainly no one-size-fits-all setting for it. Settings that have been tuned to work great over the full year turn out to be not so great when run over a more recent period such from 1st December.

In hindsight this is pretty obvious, but it highlights the fact you need to continually tweak your algorithm for current market conditions to get best results.

Anyway, I'm too late to enter but my current best for period 1st January to 21st December on MtGox with 0.55% fee is:

$1757649.52 (2630.01512 BTC)

Impressive Smiley I haven't played much with the full year performance since my k1b.. which is shy of 1.67M...
hero member
Activity: 514
Merit: 500
what I havent seen yet but seems logical is an exponential backtesting system, similar to the way ema works, so the recent trades in a back test are given more importance then the older ones. how it would work Im not exactly sure and the coding is beyond me but I am sure the results would be much more akin to what the market is currently doing.
newbie
Activity: 7
Merit: 0
I've been playing around with my own version of the Ichimoku algorithm (inspired by medji's) and am finding that there is certainly no one-size-fits-all setting for it. Settings that have been tuned to work great over the full year turn out to be not so great when run over a more recent period such from 1st December.

In hindsight this is pretty obvious, but it highlights the fact you need to continually tweak your algorithm for current market conditions to get best results.

Anyway, I'm too late to enter but my current best for period 1st January to 21st December on MtGox with 0.55% fee is:

$1757649.52 (2630.01512 BTC)
newbie
Activity: 15
Merit: 0
Starting balance — 1 January 2013: 5000 USD
Ending balance — 21 December 2013:

Medji 1108538 USD (1714.45422 BTC)
Itod 516255 USD (798.53891 BTC) (bought)

Non-contestants:

Kugutsumen I 1594742 USD (2466.41158 BTC)
Frogdog 1309577 (2025.37762 BTC)
newbie
Activity: 15
Merit: 0
Week 3 intermediate results
Kugutsumen
Balance: 1430989.95 (USD)

newbie
Activity: 28
Merit: 0
about trading, I'm invest 4 days return 100%? only here :

I have live video from them
 Smiley
full member
Activity: 220
Merit: 100
Week 3 intermediate results

itod http://cryptotrader.org/backtests/xK9hliiflmuAo5ixF
Balance: 798.53891 BTC (460756.95 USD)

ZirconiumX http://cryptotrader.org/backtests/hLWQCE1ib2kfYpS1r
Balance: 278.94306 BTC (160950.15 USD)

exuals http://cryptotrader.org/backtests/4WbpIVxxmqcEaSgCC
Balance: 376.58086 BTC (217287.16 USD)

sukiho http://cryptotrader.org/backtests/RhOj11A3ZA8PPtX2q
Balance: 875274.08 USD

Kiri11 http://cryptotrader.org/backtests/cWpHIIioI6bsv3foR
Balance: 469429.82 USD

medji http://cryptotrader.org/backtests/OEQUDQmWSWRYiEHyX
Balance: 994711.01 USD

MeatPopsicle http://cryptotrader.org/backtests/5Am6Ighnm9zYF7Ilj
Balance: 118282.47 USD

Grinny http://cryptotrader.org/backtests/WXVP1FId43h3oI9DS
Balance: 310528.61 USD

Buy'n'Hold http://cryptotrader.org/backtests/SLFrtMbfuotPruSPd
Balance: 368.06042 BTC (212370.86 USD)

The results conclusively demonstrate that Buy and Hold strategy does not seem to be a good choice for mid-term investment.
newbie
Activity: 31
Merit: 0
I (and maybe a few others too?) got this PM from frogDog, a new user who has no permission to post yet.
He asked me to post his message in this thread, so here it is:

Hey,

after all the mess kugutsumen created in the contest thread of cryptotrader, I decided to take a deep look into the strategy, and I'm pretty sure, that I found the bug.

When you take a look on http://ichimokutrader.com/elements.html#Chikou you see, that the Chikou should be backplaced, and in the http://ichimokutrader.com/signals.html page at the bottom you can see, that it should be compared with the price now.

So to state it clearly, its nothing else but comparing the actual price with the price 28 (or whatever) days ago.

Putting this into the strategy leads to following results:

https://cryptotrader.org/backtests/rjPQ3BPv54wtLbzuS


However, its not as fancy, as the results look like. Actually, they trade totally the same except on the big crash 4/10.

This lead to an advantage of around 14%, that just transfered to now. And you can reassure this, all the results kugutsumen posted, differ by the exact amount, meaning his algorithm doesnt differ much as medjis.

Thank you kugutsumen anyway, as I wouldn't have found it, without someone mentioning it.

Until today I just read in this forum, so I'm not eligble to post anything, so it would be nice, if you can post this in the contest thread as a quote.
Wow, I was quite perplexed by the claims of a bug, I thought it was in the writing of the ichimoku object (not that there isn't a bug... what the hell do I know?) not a difference in the strategy, the idea was that the chikou would give a bullish/bearish signal if it was above/below the kumo X periods past, but this looks better, thanks!
newbie
Activity: 15
Merit: 0
I (and maybe a few others too?) got this PM from frogDog, a new user who has no permission to post yet.
He asked me to post his message in this thread, so here it is:

Hey,

after all the mess kugutsumen created in the contest thread of cryptotrader, I decided to take a deep look into the strategy, and I'm pretty sure, that I found the bug.

When you take a look on http://ichimokutrader.com/elements.html#Chikou you see, that the Chikou should be backplaced, and in the http://ichimokutrader.com/signals.html page at the bottom you can see, that it should be compared with the price now.

So to state it clearly, its nothing else but comparing the actual price with the price 28 (or whatever) days ago.

Putting this into the strategy leads to following results:

https://cryptotrader.org/backtests/rjPQ3BPv54wtLbzuS


However, its not as fancy, as the results look like. Actually, they trade totally the same except on the big crash 4/10.

This lead to an advantage of around 14%, that just transfered to now. And you can reassure this, all the results kugutsumen posted, differ by the exact amount, meaning his algorithm doesnt differ much as medjis.

Thank you kugutsumen anyway, as I wouldn't have found it, without someone mentioning it.

Until today I just read in this forum, so I'm not eligble to post anything, so it would be nice, if you can post this in the contest thread as a quote.

Well good frogDog Wink

Itod, instead of insulting me, frogDog did this the right thing: RTFM & Use the source and say thank you...

2013-09-01 04:00 Simulation started. Balance: 858.00000 BTC (122069.28 USD)

Code:
Frogdog: 2013-12-13 06:00 Simulation completed. Balance: [b]1313416.44 USD[/b]
Kugu I: Simulation completed. Balance: [b]1429824.14 USD[/b]
newbie
Activity: 15
Merit: 0
Kugutsumen I
Code:
2013-01-01 00:00 Simulation started. Balance: 5000.00 USD
2013-12-12 16:00 Simulation completed. Balance: 1303663.26 USD

n.b. I am not in the contest. However Pulsecat can confirm the results.



So you're not willing to share your algo in public like we all did? :/

Did you read my post? I said three times that I will wait until round 6 and I am currently writing a clean room implementation. I am trading live with version IIa.
sr. member
Activity: 390
Merit: 250
I (and maybe a few others too?) got this PM from frogDog, a new user who has no permission to post yet.
He asked me to post his message in this thread, so here it is:

Hey,

after all the mess kugutsumen created in the contest thread of cryptotrader, I decided to take a deep look into the strategy, and I'm pretty sure, that I found the bug.

When you take a look on http://ichimokutrader.com/elements.html#Chikou you see, that the Chikou should be backplaced, and in the http://ichimokutrader.com/signals.html page at the bottom you can see, that it should be compared with the price now.

So to state it clearly, its nothing else but comparing the actual price with the price 28 (or whatever) days ago.

Putting this into the strategy leads to following results:

https://cryptotrader.org/backtests/rjPQ3BPv54wtLbzuS


However, its not as fancy, as the results look like. Actually, they trade totally the same except on the big crash 4/10.

This lead to an advantage of around 14%, that just transfered to now. And you can reassure this, all the results kugutsumen posted, differ by the exact amount, meaning his algorithm doesnt differ much as medjis.

Thank you kugutsumen anyway, as I wouldn't have found it, without someone mentioning it.

Until today I just read in this forum, so I'm not eligble to post anything, so it would be nice, if you can post this in the contest thread as a quote.
sr. member
Activity: 390
Merit: 250
Kugutsumen I
Code:
2013-01-01 00:00 Simulation started. Balance: 5000.00 USD
2013-12-12 16:00 Simulation completed. Balance: 1303663.26 USD

n.b. I am not in the contest. However Pulsecat can confirm the results.



So you're not willing to share your algo in public like we all did? :/
newbie
Activity: 15
Merit: 0
Kugutsumen I
Code:
2013-01-01 00:00 Simulation started. Balance: 5000.00 USD
2013-12-12 16:00 Simulation completed. Balance: 1303663.26 USD

n.b. I am not in the contest. However Pulsecat can confirm the results.

full member
Activity: 220
Merit: 100
Week 2 intermediate results:

itod http://cryptotrader.org/backtests/9ElJ39pdTYUI1Hmhx
Balance: 798.53891 BTC (718685.02 USD)

ZirconiumX http://cryptotrader.org/backtests/cppLTc4ksMIIOODCY
Balance: 278.94306 BTC (251048.75 USD)

sukiho http://cryptotrader.org/backtests/xWWatTMRigLg4WPhW
Balance: 994526.45 USD

exuals http://cryptotrader.org/backtests/jxyrDOX2soTVVhWnF
Balance: 376.58086 BTC (338922.78 USD)

Kiri11 http://cryptotrader.org/backtests/UgHP8CUG4n2QepSIm
Balance: 664.82530 BTC (598342.77 USD)

MeatPopsicle http://cryptotrader.org/backtests/rbg7tgSg4AvGJq7TI
Balance: 154105.46 USD

medji http://cryptotrader.org/backtests/9PwJqAGU1oy6tQ4ch
Balance: 1111906.73 USD

Grinny http://cryptotrader.org/backtests/0UYMqa7adTEuBPBib
Balance: 334532.63 USD

Buy'n'Hold http://cryptotrader.org/backtests/rtxY6iebkMHud4h7q
Balance: 368.06042 BTC (331254.38 USD)
member
Activity: 66
Merit: 10
If people continue sharing great ideas ,the service will continue to get better. That's true for anything, but it should be especially true for the bitcoin community : )) No need to fight 
newbie
Activity: 15
Merit: 0
Quote
If you are claiming there's a bug in Ichimoku example in Backtest->Featured part of the site, which are there for everyone to use, and you don't want to point which bug is it - that's really lame. Don't do such a thing.

I think that the Ichimoku example has a small bug in it but that is based on my initial reading of the docs at ichimokutrader.com. I'll probably share my fix when round 6 opens. It will give me time to research my theory further, I don't see how this is lame.

Well, it's a peace of opensource code you've got for free to work on. Holding back findings about bugs so everyone can continue to work on that same peace of code is lame in my book. You don't have to offer your fixes to everyone, but you should describe the bug.

Why should I do that? Why do you need to insult me? I've already stated that I will probably participate in round 6 and if I do, I will have to release the code.

About open source:

Code:
Nobody should be restricted by the software they use. There are four freedoms that every user should have:

the freedom to use the software for any purpose,
the freedom to change the software to suit your needs,
the freedom to share the software with your friends and neighbors, and
the freedom to share the changes you make.

Hint: Freedom is the keyword...

Until recently, my real life job was to find hard-to-find bugs and vulnerabilities in software and systems, mostly banking software (core banking, treasury systems, mobile banking, online banking...)  We get paid a lot of money to do that. Open source or closed source software, it doesn't matter, we still get paid.

Even the restrictive GPL doesn't force you to publish any changes unless you distribute compiled code.

In copyright law, a derivative work is an expressive creation that includes major, copyright-protected elements of an original, previously created first work (the underlying work). The derivative work become a second, separate work independent in form from the first. The transformation, modification or adaption of the work must be substantial and bear its author's personality to be original and thus protected by copyright.

In the particular case of cryptotrader.org, I actually paid Alex to use his service and infrastructure. Regarding Medji or the sample code, I don't know their respective history, where the source code comes from and who wrote it.  What I know is that Goichi Hosoda, the creator of Ichimoku Kinkō Hyō is not getting anything from Alex or Medji or any of the traders and software engineers who uses his methology.

If you had just spent an hour reading up on Ichimoku and took a critical look at the code, you would have noticed the bug right away because it is so obvious!! I even thought it was left there as an exercise. (and in the old days, released exploit code would includes bugs and typos on purpose to make life harder for script kiddies and what we would call lamers...)
Pages:
Jump to: