rpietila Wall Observer - the Quality TA Thread ;) - page 207.

rpietila

donator

Activity: 1722

Merit: 1036

Quote from: darlidada on May 10, 2014, 02:23:47 PM

I put some money on an exchange the day bitcoin hits 360$ and i'm still waiting a window to buy.... I dont think I am the only one. Now thats its back at 450, I wish I had bought 5 days ago. I'm sure I'm not the only one.

What shall I do rpietila?

Depends on how many % of your liquid worth you already have invested in bitcoins. If you are already fully invested, it might not make sense to buy now because the downtrend may continue and you don't want to end up selling your coins at anywhere near these prices. So it's good to have some cash reserve.

But if you are moderately invested (<50% of total assets), and want to buy more, this is a good time. Although if you are investing long-term, you can as well wait for the breach of 500. It does not really matter because the selling price is so much higher. The important thing is to buy before the next big runup.

If we are talking about short-term only, I refuse to give any advice since I don't know your trading strategy.

darlidada

hero member

Activity: 723

Merit: 503

I put some money on an exchange the day bitcoin hits 360$ and i'm still waiting a window to buy.... I dont think I am the only one. Now thats its back at 450, I wish I had bought 5 days ago. I'm sure I'm not the only one.

What shall I do rpietila?

rpietila

donator

Activity: 1722

Merit: 1036

Quote from: rpietila on May 05, 2014, 09:31:13 AM

Quote from: rpietila on April 30, 2014, 02:30:17 AM

Quote from: rpietila on April 26, 2014, 10:31:30 AM

Quote from: rpietila on April 24, 2014, 06:31:10 AM

Quick TA update (at $0.481):
- 6H candle color/volume: not much volume during this week, conclusion: none/(historically: bearish)
- Bid/ask strengh at market (Bitstamp): slippage to sell 5k: $0.040, slippage to buy: $0.043, conclusion: both sides have built up since last time
- Trendline comparison: we are now at -0.355 log units. The trendline is at $1.090 and rising $0.008 per day, conclusion: rock bottom (note: it is not necessary that 'rock bottom' will change until the parabolic uptrend starts, because the trendline is itself rising)
- Sentiment: n/a
- Prognosis: shortest term unsecure, mid-term not sure if reversal seems confirmed, long-term buy zone

Quick TA update (at $0.460):
- 6H candle color/volume: red candles are taller, conclusion: bearish
- Bid/ask strengh at market (Bitstamp): slippage to sell 5k: $0.040, slippage to buy: $0.056, conclusion: easier to move to the upside
- Trendline comparison: we are now at -0.365 log units. The trendline is at $1.060 and rising $0.007 per day, conclusion: rock bottom (the 12-month range of the trendline is -0.404...+0.518)
- Sentiment: fearish, bearish, conclusion: less downside than generally feared
- Prognosis: shortest term don't know, mid-term not sure if reversal seems confirmed, long-term buy zone

ADD: If sentiment is bearish, it indicated that the price will surprise to the upside, and vice versa.

Quick TA update (at $0.450):
- 6H candle color/volume: GREEN rules for last 5 days, although short, conclusion: slightly bullish
- Bid/ask strengh at market (Bitstamp): slippage to sell 5k: $0.037, slippage to buy: $0.049, conclusion: pressure to the downside, surprise to the upside (when one side has lower figure than the other, it means that price has moved to that direction, eroding the bids/asks - if there is a reversal, it can move quickly to the other side with less resistance)
- Trendline comparison: we are now at -0.384 log units. The trendline is at $1.090 and rising $0.008 per day, conclusion: rock bottom (the 12-month range of the trendline is -0.404...+0.518)
- Sentiment: weak
- Prognosis: shortest term don't know, mid-term not sure if reversal seems confirmed, long-term (3+ months) buy zone

Quick TA update (at $0.429):
- 6H candle color/volume: very little volume for last few days, conclusion: historically bearish
- Bid/ask strengh at market (Bitstamp): slippage to sell 5k: $0.029, slippage to buy: $0.045, conclusion: pressure to the downside, surprise to the upside (when one side has lower figure than the other, it means that price has moved to that direction, eroding the bids/asks - if there is a reversal, it can move quickly to the other side with less resistance)
- Trendline comparison: we are now at -0.420 log units. The trendline is at $1.128 and rising $0.008 per day, conclusion: rock bottom (the 12-month range of the trendline is -0.420...+0.518)
- Sentiment: n/a
- Prognosis: short term risk for downside, mid-term we are still in a downtrend, long-term (3+ months) strong buy zone

Quick TA update (at $0.449):
- 6H candle color/volume: greens rule on quite little volume, conclusion: slightly bullish
- Bid/ask strength at market (Bitstamp): slippage to sell 5k: $0.029, slippage to buy: $0.045, conclusion: buy pressure has created potential for further upside
- Trendline comparison: we are now at -0.413 log units. The trendline is at $1.161 and rising $0.008 per day, conclusion: rock bottom (the 12-month range of the trendline is -0.420...+0.518)
- Sentiment: hopeful
- Prognosis: short term likely upside, mid-term breakout from the downtrend not confirmed, long-term (3+ months) strong buy zone

The backtested trading signals are now SHORT. The signal changes to LONG in the event of a large enough move up with large enough volume that beats the volume of a long enough time before it.

rpietila

donator

Activity: 1722

Merit: 1036

Quote from: Trolololo on May 08, 2014, 08:52:06 PM

Quote from: rpietila on May 06, 2014, 02:42:49 AM

The candle is locally-longest if it is longer than 3 candles in both sides. The locally-longest candle dominates the said candles in either side of it and they don't count no matter the color.

You can't know the 3 candles in the future.
So you have to wait 3 periods to determine wether a candle is locally dominant.

Have you simulated what's the rentability of your strategy if you place the buy/sell order 3 periods later than the locally dominant candle?

The script you quoted was based on my earlier, "soft" way of assessing whether we are in an uptrend or not. It was criticized because of unaccountability (by the very people who never divulge anything they have actually done, just claiming to have been, and being, correct all the time which is kind of annoying).

So I wanted to write an algorithm that would work in a similar way but actually be implementable:

Quote from: rpietila on May 08, 2014, 04:30:36 PM

IF "the last row volume is at least A" &
"the highest volume in the last B rows is to the same direction as the last volume" &
"the last change is at least C" &
"the new signal is to the opposite direction as the current signal",
THEN give a new signal.

- The parameters A, B, C can be fully customized. As you change them, you see in real time what signals the system would have given and when, how many total signals (too many leads to increase in fees), and what is the end result in $ or bitcoins in this time period. You can try several values per minute, and then zoom in to the interesting parts of the data, such as the flashcrashes.

It is not looking into the future, but it places a lag on the trend changes because the volume towards the new trend direction has to be greater than the volume in the old direction, in any of the previous B time periods.

I am hesitant to implement this because it is based on volume, which is so easy to manipulate. Also this one seems to do terribly once the rise starts. (If you can know the exact high and low points of the downtrend, even more than 3-4x your bitcoins can easily be achieved so trading algo is not needed Wink

)

Price is not considered when it gives the signals, haha funny. Blind monkey.

Trolololo

sr. member

Activity: 263

Merit: 280

Quote from: rpietila on May 06, 2014, 02:42:49 AM

(...)
The candle is locally-longest if it is longer than 3 candles in both sides. The locally-longest candle dominates the said candles in either side of it and they don't count no matter the color.
(...)

Interesting strategy, Risto.
But I have one doubt:

You can't know the 3 candles in the future.
So you have to wait 3 periods to determine wether a candle is locally dominant.

Have you simulated what's the rentability of your strategy if you place the buy/sell order 3 periods later than the locally dominant candle?

SlipperySlope

hero member

Activity: 686

Merit: 501

Stephen Reed

Regarding trading bots and systems for Bitcoin, I am very much hopeful that Chinese exchanges while expanding outside the mainland, bring with them the high volume and good order depth that resulted from their competitive policies of zero-fee trading.

I would like to see OKCoin or Huobi host the BTC/USD currency pair with zero fees.

It pains me to add up the round trip commissions that I paid Mt. Gox. I wrote a Mt.Gox bot but never deployed it. Market making bots, e.g. Volume Weighted Average Price algorithm, use realtime technical indicators to make frequent trades near the spread. Exchange commissions make operating such bots unprofitable.

rpietila

donator

Activity: 1722

Merit: 1036

Quote from: donut on May 08, 2014, 05:23:20 PM

I will think about it if you post a backtest from April 11 to Sept 26, 2013 with the same parameters you used for the current bubble deflation period, aka no overfitting.

Very same parameters used, which may be near optimal (but likely are not, especially volume threshold was fixed figure and BS volume was not generally that big those days..) :

Code:

10.4.2013 18:00 130,00 -1 13 000 0,00
16.4.2013 14:00 74,00 1 76 % 0 175,68
19.4.2013 8:00 120,16 -1 21 109 0,00
3.5.2013 12:00 90,10 1 33 % 0 234,29
6.5.2013 20:00 110,66 -1 25 926 0,00
4.7.2013 6:00 76,99 1 44 % 0 336,75 <--- !!!
5.7.2013 16:00 68,50 -1 23 067 0,00
10.7.2013 18:00 79,41 1 -14 % 0 290,48
18.7.2013 6:00 79,22 -1 23 012 0,00
19.8.2013 6:00 102,60 1 -23 % 0 224,29
6.9.2013 18:00 117,05 -1 26 253 0,00
9.9.2013 14:00 122,35 1 -4 % 0 214,57

Notable is that it has generated 3.4x the number of bitcoins during the 3-mo downtrend despite selling only at $130 in the start. But at the instant the price starts to go up, it has lost quite many coins.

windjc

legendary

Activity: 2156

Merit: 1070

Quote from: rpietila on May 08, 2014, 04:30:36 PM

Quote from: rpietila on May 08, 2014, 03:42:56 PM

The configurable excelsheet is for sale for BTC5, including tutorial in skype chat.

This is actually a good deal, I will open it up in case somebody is interested.

The sheet works as follows:
- You can copypaste any data from Bitcoincharts.
- The system gives buy and sell signals according to the following rule:
IF "the last row volume is at least A" & "the highest volume in the last B rows is to the same direction as the last volume" & "the last change is at least C" & "the new signal is to the opposite direction as the current signal", THEN give a new signal.
- The parameters A, B, C can be fully customized. As you change them, you see in real time what signals the system would have given and when, how many total signals (too many leads to increase in fees), and what is the end result in $ or bitcoins in this time period. You can try several values per minute, and then zoom in to the interesting parts of the data, such as the flashcrashes.

So yes, I am giving all this work for (almost) free.

I'll do this. Pay me 5 BTC and I'll take the chart and trade it for 6 months and give you 20% of the profits. Deal? KTHXBYE

donut

sr. member

Activity: 248

Merit: 252

Quote from: rpietila on May 08, 2014, 04:30:36 PM

Quote from: rpietila on May 08, 2014, 03:42:56 PM

The configurable excelsheet is for sale for BTC5, including tutorial in skype chat.

This is actually a good deal, I will open it up in case somebody is interested.

The sheet works as follows:
- You can copypaste any data from Bitcoincharts.
- The system gives buy and sell signals according to the following rule:
IF "the last row volume is at least A" & "the highest volume in the last B rows is to the same direction as the last volume" & "the last change is at least C" & "the new signal is to the opposite direction as the current signal", THEN give a new signal.
- The parameters A, B, C can be fully customized. As you change them, you see in real time what signals the system would have given and when, how many total signals (too many leads to increase in fees), and what is the end result in $ or bitcoins in this time period. You can try several values per minute, and then zoom in to the interesting parts of the data, such as the flashcrashes.

So yes, I am giving all this work for (almost) free.

I will think about it if you post a backtest from April 11 to Sept 26, 2013 with the same parameters you used for the current bubble deflation period, aka no overfitting.

rpietila

donator

Activity: 1722

Merit: 1036

Quote from: rpietila on May 08, 2014, 03:42:56 PM

The configurable excelsheet is for sale for BTC5, including tutorial in skype chat.

This is actually a good deal, I will open it up in case somebody is interested.

The sheet works as follows:
- You can copypaste any data from Bitcoincharts.
- The system gives buy and sell signals according to the following rule:
IF "the last row volume is at least A" & "the highest volume in the last B rows is to the same direction as the last volume" & "the last change is at least C" & "the new signal is to the opposite direction as the current signal", THEN give a new signal.
- The parameters A, B, C can be fully customized. As you change them, you see in real time what signals the system would have given and when, how many total signals (too many leads to increase in fees), and what is the end result in $ or bitcoins in this time period. You can try several values per minute, and then zoom in to the interesting parts of the data, such as the flashcrashes.

So yes, I am giving all this work for (almost) free.

rpietila

donator

Activity: 1722

Merit: 1036

The thread shows a poster example how one system would have performed during the downtrend of the past few months. The "track record" is impressive but we have to remember that I started from the data and honed the parameters with trial and error method to give the best results. In real world, spotting the downtrend correctly, guessing the parameters upfront, and executing without slippage would have diminished returns greatly.

I cannot sell my own signals because Bitstamp's average volume is BTC764 per hour. The system gives signals based on the last 2 hours market data and assumes that you can execute at the closing price. The signals are always in direction of the trend, so you will have to buy/sell quickly to avoid slippage, and your own actions cause more slippage. If I would use this, I want to allocate as many coins to it as possible, and that is less than I wanted. Sad

The configurable excelsheet is for sale for BTC5, including tutorial in skype chat.

MahaRamana

sr. member

Activity: 338

Merit: 250

Quote from: rpietila on May 08, 2014, 12:51:19 PM

Quote from: hbee on May 08, 2014, 12:09:09 PM

Quote from: rpietila on May 08, 2014, 10:29:36 AM

Newest automatic volume-based trading system was backtested with the data from 2013-12-1...2014-5-8. The parameters that gave the so far optimal result were:

- 2H data, buy/sell signal comes from a high-volume change into that direction
- Quite large minimum trigger volume for the last 2H (less than 15% of 2H periods qualify, this is also good for high-volume trading)
- Quite large minimum % change for the last 2H (less than 15% qualify)

I won't disclose the exact parameters sorry (because this is starting to look too good Wink

) but this one has so far returned BTC427 for the investment of BTC100 in 2013-12-1, less than 6 months ago. Number of single trades (not pairs) was 19.

ADD: Current record is 442 bitcoins from 100 bitcoins with 17 trades. That's 9.1% per trade. Cool

Intriguing numbers, what software are you using for this analysis?

MS Excel.

The record trading algo would have performed as follows since 2013-12-1:

Code:

TIME OF TRADE RATE SELL/BUY B GAIN USD BTC
1.12.2013 4:00 1024,00 -1 102 400 0,00
7.12.2013 2:00 748,99 1 37 % 0 136,72
11.12.2013 10:0 852,50 -1 116 552 0,00
19.12.2013 8:00 605,00 1 41 % 0 192,65
20.12.2013 22:0 610,00 -1 117 515 0,00
22.12.2013 8:00 652,86 1 -7 % 0 180,00
7.1.2014 20:00 823,00 -1 152 030 0,00
14.2.2014 2:00 595,00 1 38 % 0 255,51
20.2.2014 6:00 601,10 -1 153 589 0,00
22.2.2014 16:00 603,98 1 0 % 0 254,29
24.2.2014 2:00 574,97 -1 146 211 0,00
26.2.2014 6:00 594,98 1 -3 % 0 245,74
4.3.2014 16:00 675,61 -1 166 026 0,00
1.4.2014 6:00 481,01 1 40 % 0 345,16
2.4.2014 12:00 469,95 -1 162 208 0,00
14.4.2014 4:00 427,50 1 10 % 0 379,43
16.4.2014 8:00 515,00 -1 195 409 0,00
8.5.2014 16:00 442,50 1 16 % 0 441,60

Looks nice, are you selling it ?

rpietila

donator

Activity: 1722

Merit: 1036

Quote from: hbee on May 08, 2014, 12:09:09 PM

Quote from: rpietila on May 08, 2014, 10:29:36 AM

Newest automatic volume-based trading system was backtested with the data from 2013-12-1...2014-5-8. The parameters that gave the so far optimal result were:

- 2H data, buy/sell signal comes from a high-volume change into that direction
- Quite large minimum trigger volume for the last 2H (less than 15% of 2H periods qualify, this is also good for high-volume trading)
- Quite large minimum % change for the last 2H (less than 15% qualify)

I won't disclose the exact parameters sorry (because this is starting to look too good Wink

) but this one has so far returned BTC427 for the investment of BTC100 in 2013-12-1, less than 6 months ago. Number of single trades (not pairs) was 19.

ADD: Current record is 442 bitcoins from 100 bitcoins with 17 trades. That's 9.1% per trade. Cool

Intriguing numbers, what software are you using for this analysis?

MS Excel.

The record trading algo would have performed as follows since 2013-12-1:

Code:

TIME OF TRADE RATE S/BUY B GAIN USD BTC
1.12.2013 4:00 1024,00 -1 102 400 0,00
7.12.2013 2:00 748,99 1 37 % 0 136,72
11.12.2013 10:0 852,50 -1 116 552 0,00
19.12.2013 8:00 605,00 1 41 % 0 192,65
20.12.2013 22:0 610,00 -1 117 515 0,00
22.12.2013 8:00 652,86 1 -7 % 0 180,00
7.1.2014 20:00 823,00 -1 152 030 0,00
14.2.2014 2:00 595,00 1 38 % 0 255,51
20.2.2014 6:00 601,10 -1 153 589 0,00
22.2.2014 16:00 603,98 1 0 % 0 254,29
24.2.2014 2:00 574,97 -1 146 211 0,00
26.2.2014 6:00 594,98 1 -3 % 0 245,74
4.3.2014 16:00 675,61 -1 166 026 0,00
1.4.2014 6:00 481,01 1 40 % 0 345,16
2.4.2014 12:00 469,95 -1 162 208 0,00
14.4.2014 4:00 427,50 1 10 % 0 379,43
16.4.2014 8:00 515,00 -1 195 409 0,00
8.5.2014 16:00 442,50 1 16 % 0 441,60

So guys, with only 17 trades, performed after the triggers (perhaps BTC500 could have been used without affecting the market too much - even thousands with some planning, delay, and slippage), you would have 4.5x'd your bitcoins or made almost double your dollars, in a falling market.

I also checked the same rules in a rising market. They did not work, I could barely make it return positive bitcoins over 1 year. The problem is that in the rising market the selloffs come viciously and up it goes with stable medium volume. So this interprets the periodic selloffs as the important events and loses out on runups.

hbee

newbie

Activity: 4

Merit: 0

Quote from: rpietila on May 08, 2014, 10:29:36 AM

Newest automatic volume-based trading system was backtested with the data from 2013-12-1...2014-5-8. The parameters that gave the so far optimal result were:

- 2H data, buy/sell signal comes from a high-volume change into that direction
- Quite large minimum trigger volume for the last 2H (less than 15% of 2H periods qualify, this is also good for high-volume trading)
- Quite large minimum % change for the last 2H (less than 15% qualify)

I won't disclose the exact parameters sorry (because this is starting to look too good Wink

) but this one has so far returned BTC427 for the investment of BTC100 in 2013-12-1, less than 6 months ago. Number of single trades (not pairs) was 19.

ADD: Current record is 442 bitcoins from 100 bitcoins with 17 trades. That's 9.1% per trade. Cool

Intriguing numbers, what software are you using for this analysis?

rpietila

donator

Activity: 1722

Merit: 1036

Quote from: holomen on May 08, 2014, 11:21:17 AM

Congratulations Risto! You Rock, how long are you trading since to know so much?

Sounds like a nice rocket science to me;)

P.s. Where should I start learning how to make that ATS, could you tell few books or links ? Wink

I'm serious.

I started online stock trading in 1998-99, before the Nasdaq bubble.

I think bitcoin is so special that the books are of limited benefit only. Hands-on analysis and experience is what rocks.

holomen

newbie

Activity: 54

Merit: 0

Quote from: rpietila on May 08, 2014, 10:29:36 AM

Newest automatic volume-based trading system was backtested with the data from 2013-12-1...2014-5-8. The parameters that gave the so far optimal result were:

- 2H data, buy/sell signal comes from a high-volume change into that direction
- Quite large minimum trigger volume for the last 2H (less than 15% of 2H periods qualify, this is also good for high-volume trading)
- Quite large minimum % change for the last 2H (less than 15% qualify)

I won't disclose the exact parameters sorry (because this is starting to look too good Wink

) but this one has so far returned BTC427 for the investment of BTC100 in 2013-12-1, less than 6 months ago. Number of single trades (not pairs) was 19.

ADD: Current record is 442 bitcoins from 100 bitcoins with 17 trades. That's 9.1% per trade. Cool

Congratulations Risto! You Rock, how long are you trading since to know so much?

Sounds like a nice rocket science to me;)

P.s. Where should I start learning how to make that ATS, could you tell few books or links ? Wink

I'm serious.

rpietila

donator

Activity: 1722

Merit: 1036

Newest automatic volume-based trading system was backtested with the data from 2013-12-1...2014-5-8. The parameters that gave the so far optimal result were:

- 2H data, buy/sell signal comes from a high-volume change into that direction
- Quite large minimum trigger volume for the last 2H (less than 15% of 2H periods qualify, this is also good for high-volume trading)
- Quite large minimum % change for the last 2H (less than 15% qualify)

I won't disclose the exact parameters sorry (because this is starting to look too good Wink

) but this one has so far returned BTC427 for the investment of BTC100 in 2013-12-1, less than 6 months ago. Number of single trades (not pairs) was 19.

ADD: Current record is 442 bitcoins from 100 bitcoins with 17 trades. That's 9.1% per trade. Cool

rpietila

donator

Activity: 1722

Merit: 1036

OK, fine. There is one month in the year when I am in tune with the Bitcoin universe, and it is the month when the price rises by a factor of 5-10. For 2014, it is yet to come..

Meanwhile I was analysing the previous trading system more and the period between 2014-2-1 and 2014-3-13 (40 days) still returned 29% more bitcoins (with 2*9 trades) with a set of very simple rules that I just pulled from the hat. By optimizing all the parameters I am sure it will return better.

TERA

hero member

Activity: 728

Merit: 500

Where do you get the number 2 years? I have this chart hitting a bottom in 3 months and breaking ATH in 1.5 years. Also, it doesn't require that bitcoin adoption stop. It only requires that the supply is greater than the demand. Right now the demand has decreased by an unknown factor due to the chinese shadow currency traders in the past, and the supply has increased by millions of coins between mtgox, china, the government, silk road, etc. All this adoption would be the reason why we capitulate at 270 and not any lower.

rpietila

donator

Activity: 1722

Merit: 1036

It may happen, but only if the exponential, viral, increase of knowledge considering Bitcoin's (and cryptotechnology's) merits would immediately stall and nobody would talk about it for 2 years, nor would there be any maturation of the thinking in the heads of those who already know about the coin. Also all the bitcoin companies and projects in pipeline need to shut down.

=> Summary shutdown of the Internet and forced lobotomy for all males under 60.

Quote from: TERA on May 08, 2014, 06:50:17 AM

Perfect

But hardly realistic.

Topic: rpietila Wall Observer - the Quality TA Thread ;) - page 207. (Read 907212 times)