Krypto-trading-bot: Missed opportunities

[AZm1n]

Hello All,
I've been testing the bot for about a week in LTC-USD market, I played with every settings to get familiar with it. The good news with this bot is at least you don't lose the money. but at the same time it won't grow it either.

The issue here that the bot misses a lot of good sell/buy opportunities, for example the price dropped from 51.30 to 50:30 and the bot kept dropping the quoted price and move it away until the price finally started to rise again and instead of buying at 50.30, it bought it 50:89!

I tried with the lowest width possible, different Stdev settings, BB on and off, different TBP/PDiv, but they are all the same at the end

From my week observation with whatever settings I choose, the bot start buying when the price is in decline or about to decline.
And selling quickly when the price start to rise fractionally above the buy price.
So the margin at the end is very slim.
I just get out break even every time and I feel like working for the exchange! I don't know but apparently the bot work in reverse.

Looking at the chart manually, I easy made 10$ in one hour using two direct trades on the exchange, with the bot I really have to wait longer and the return is not big at the end

I'm not asking for a strategy, (read quite a few) I but want to know how to make bot more smart/efficient to such price drops and falls.
any general tips/tricks? sometime it missies an opportunity every hour or so for easy .5$ profit.

I really did my homework and read a lot about trading strategies but the point above is just killing me, other wise this will be an awesome bot.

I really want to send some chocolate milk but to do that the cow must produce some milk first and then i'll add the chocolate :-)

Thanks.

[AZm1n]

To make it little bit simpler, The bot sells when the price is trending up and buys when the price is trending down, even though according to the Tbp and Pdiv i'm not out of position :-)

Example:
Price : 51.30 => 51.10 => 50.95 => 50.80 =>....50.30
Bot Quote: => 51.02 => 50.85 => 50.70..............50:10

Price start to rise in reverse and the bot now also tries to catch the new price to no avail.
Price then reached 51.60 =>Bot started to sell
Price started to drop to 51.05 => The bot will buy here while price trend is still clearly pointing down.
So then I have to wait for the price to rise a little above 51:05 and then the Bot will again begin sell.
so while it doesn't lose the money as a whole at whatever price point of the coin at any given time simply fiat remain the same.

Thanks again.

[ctubio]

i think what you need is a bigger width, because a small width will cause fast buy/sells asap at the beginning of the drop/rise; instead if you set a bigger width it will at least make bigger margins avoiding the small trades in the market and only make matched orders on big market trades (still without caring if these are drop/rises beginnings or not)

if you want somehow to tell the bot "what is a beginning or end of a drop/rise" the ony way is to define it playing with apMode/ewma/stdev values; if these are Manual or Off, then the bot will be a pure ping-pong bot; otherwise if you active and configure them, you can take profit (or looses) also from market-making bot.

[Camille92]

Hi both or you,

@AZm1n I have a few things to tell you about your quest of a profitable strategy.
The bot comes "as is" and will just automate your strategies so if your strategy is wrong, it's going to lose money or not make a profit.

Sure you can always make a profit trading directly on the exchange, but you have to realize that K has access and can manage WAY less information than a human brain and works with a limited amount of data available.
If you have a consistent and good strategy to share, we would be all glad to listen to it and incorporate it if possible. :)

The strong advantage of the bot in comparison of a human is that it is 24/7 watching the market and can react way faster than any human being.

Now if we go back to the strategies available, with the bot you have 2 different aspects you need to consider to make a profit.

1) Market making: It's in a way working for the exchange (or at least the market). You provide liquidity to the market in exchange for a small profit. (Sell-Buy-fees). It's very short term.

The settings related to market making are: Mode, Ping&PongAt, SOP, Size, Ping&Pongwidth, Fv, trades /sec, ewma, stdev, BB.

2) Technical Analysis: It's a try a guessing if the price is going up or down in the midterm.

The settings related to that are: APmode, the ewma values, sensibility and pdiv.

APR is in the middle of the two strategies.

For a good market making strategy, you have to look in your market what is the average size of an order. How deep in the order book does it go? How often? How long are the small movements?

Your goal here will be only to buy and sell as quickly as possible but making a consistent profit each time. Go to trading view and to the exchange, look at the chart and analyze what are the pattern at 1mn.

If you see that the price fluctuates by $1 every few minutes, try something with ping pong 0.8 widths or put ewma protection with 5 periods, or stdev with 300 periods and factor of one.

Do different experiments but if you can't consistently buy lower and sell higher then revise your strategy.

Then you have the Technical analysis, go again to your chart in trading view and try to look for a couple of EWMA that work well in following the trade without making too many mistakes. Then put it in your K settings. Try to look at how much they cross before you have a true signal and use that as sensibility.

Then look also how much you're making with the first strategy to see what would be your minimum pdiv possible without harming the market making strategy.

And lastly remember, we're maker here so if no one SELLS you some LTC, you're not going to buy that's why you bought at 50.89 and not 50.3, because no one was selling to you at that price. https://github.com/ctubio/tribeca/issues/100 this update should help with that :)

[AZm1n]

Thanks to both of you.
I'll try tweaking the settings again and see how it goes.
Will let you know.

Thanks again.

[Camille92]

Hello @AZm1n,

I hope you've been progressing, the last few days have been super hard to trade so don't worry if you're at a loss.

I just come to show you what you must see when you find a good market making strat:
Look how sells are consistently (but not always) way above than buys.

Best!

[AZm1n]

Thanks Camille, appreciate the follow-up.

actually yesterday I assume it was supposed to be a good day for trading :), but amazingly I still get out break even again :(
here is an image of my sells and buys for the last 6 hours, lots of buys/sells but on the edge, so the margin is relatively small.

But yesterday the market was fluctuating over 4$, from -42$ to +46$ and it did that multiple times, I didn't do any manual trades as I was focusing trying to make the bot do that for me but unfortunately not.

What's really killing is when is the decision is made to either do a buy or a sell?
The bot amazingly keep up at the price trend but then fail to either sell or buy at the right price miserably (or my settings).
The prices yesterday dropping/rising more than 1.5$ every one hour or close.
The bot don't the price at the bottom (or on the rise), and doesn't sell when the price in decline (or about to).

I tried and patiently observed different EWMA settings, sensibility, modes, playing with sizes, but the result still the same.

If i set a long width of .8, the bot simply can stay for more than 6 hours (or even longer), with no sells or buys, although there was good chances for buying and making more than 1$ profit per coin.

Then if i set a narrow width of .3, the bot just starting to do the market making but with not much money generated.

Again I look at the chart I do good decisions, but like you have said previously having that automated would be awesome.

I tried relaxed and aggressive EWMAs (10, 200, 100), (100, 50, 200) along other combinations.
played with sensibility, turned off STDEV and relied on 5 EWMAs.
different modes (primarily AK47 and Ping Pong)

but whatever the settings I choose I end up with the same result, I'm playing with about 27 Coins and the bot did around 250 trades since 3 days and I'm amazingly still ~27 to 28 ~Coins, while I expect If the bot (or my settings) was doing the right thing I was expecting a lot of money as if I'm doing the trade manually.

I still didn't give up completely yet, I'll try to play again for couple more days and see how far it goes.
an explanation of the logic of the bot to buy/sell and the way it fluctuates the price would be appreciated.

Simply I don't want the price to rise from 44 to 45 and the bot still keeping increasing the ask until the momentum changes and long gone.

I can see clearly the opportunity this bot offer but so far not able to use it. :(

Thanks again.

[Camille92]

Hi there!

So giving the info I get I can advise you to:

Look for a small profit if you're on ltc, for instance, my settings are at a pong width of 0.045, and I make a LOTS of trades. (Note that I don't pay maker fees on Bitfi now). My goal is not to make $1 per coin traded but just a few cents. To compensate, I try to trade hundreds of coins. If I remember well I trade around 30 times the size of my stack per day.
The chart I sent you is about 15mn of trading.

Now for your ewma settings, I'm not a big fan of the Long, Medium, Short because short can really send too many false signal info (I know @ctubio it was my idea). I still think there is something to do here but not in the current form.

For what it's about finding tops and bottoms there is no way to do that in K's current form. Maybe in the future, we could try to have that, either by the "securities" I developed before or by using a mix of vix / stocha ...

What you can do basically is -> Following a trend with the ewma.
Doing a lot of trades for a small profit in market making!

I think that you have to forget to try to copy "human trading" with this bot and use what it can do -> ultra fast and 24/7. And think about it, 200 trades at 1c profit per coin is better than 1 trade at 1 usd profit per coin :)

I hope it helps!

[AZm1n]

Thanks Camille,
I think I got you're point and you're absolutely right, I'm trying to use the bot as human and this seem to be the issue here :)

Let me try to change the strategy for market making one and see what results will get, will keep you posted.
Thanks again and appreciated.

[AZm1n]

Hi Camille,
I honestly still struggling to get the bot to a stable profitable condition Camille.
I still don't get the semantics of the decision to sell or buy.
The problem is that the bot sells either very fast at the beginning of a momentum, or wait for the decline and sells way below higher price and the same also happens for the buys.

I tried manual, LS, LMS and they all produce the same behavior.
Width .1 to .3 and pong width of .365 to .1
STDEV for two minutes and factor of .2 which I believe should make it more sensitive to market.

Few minutes ago the price of LTC was 42.72, The bot purchased it at 42.32.
Now it want to sell it at 41.90!!!, I know it want to to sell later at higher price, but in this situation it missed the 42.72 price as well selling ~42c lower than it what originally purchased.
That's why I see all greens in AK47 mode but when I do the math, the numbers doesn't add up :-)
I'm using the AK47 mode.

I also tried different Pdivs/TBP including crazy ones, but the behavior I described above remained the consistent behavior.
I won't blame the bot, because I think it knows what's going on, but honestly what I feel that it do the sell while it supposed to do the buy and vice versa :)

Thanks again.

[Camille92]

hey @AZm1n yep it's because buying / selling is not based on previous buy / sell price ( a part in PingPong), but rather if the bot sees it at a good opportunity to buy now and sell later higher.

You're always going to lose money in market movement, a part from the big ones where TBP and ewma are going to make you surf the wave.

For market making you should look for the spread and use that as your target profit. The best market for a maker is when it's totally flat.

And for small movements like the one you're describing no one is able to guess them haha :)

Again really make a difference in your mind between the market making part of the bot and the technical analysis one.

Market making is blind and care only about the spread and the order book :)

If you really struggle market making, put pdiv = 0 and only play with the ewma :)

[AZm1n]

Thanks Camille,
Yes I tried the market making when the market is flat, buying and selling was working but based on a slow margin, so have to compensate that with the size.

The problem is that when there is a big market move, the decision made by the bot is incorrect IMHO.
I.E. There is a big decline > Bot quickly buy Coins and then wait and rebuy again at much higher price.
There is a big Move => Bot sells quickly and resell at much lower price.

This is actually killing, because whatever money being made is getting easily lost and then i end up with the same amount of fiat currency again.

That's what I'm saying that whatever good things the bot do, it immediately gone when there is a move upward or downward.
This happened with LS/LMS and manual.

If the bot reverse the decision when market moves fast. I think it would've made a lot of money.
Why I'm saying that, again because this is the only consistent behavior I noticed while experimenting several settings :)

[AZm1n]

I think it could be a problem with my sell/buy size.
I'm currently experimenting with ~30 Coins and setting buy/sell size at 6 each.
Maybe this size fits my main human perception of this (i.e sell at high margin)
while in Market Making I should sell much less (depend on my full stack), this at least should make me in the game for longer period and sustain these bug price swings.

Will try that and tweak the PDiv further, is there any rule of thumbs for PDiv in contrast to the stack size?

Thanks for all the help.

[AZm1n]

Camille,
I have to apologize to you, my approach was totally wrong!
Now I understand how K works :-)
Now let me do some more correct experiments over the coming two to three days and let you know the outcome.

Thanks for help answering all the questions.

[AZm1n]

Hey Camille,
Sorry for the delay coming back with some results, I wanted to to do the proper testing.
Unfortunately still the bot didn't live up to my expectations :(
very small profit being made, I'm able to get K to trade for profit, but the margin is very small.
I get about .2 to .3 of profit per trade using AK47 and I can't get the bot to do much trade even when market is moving fast upward/downward.
The way by the way AK47 work doesn't make much sense to me.
If K sold for loss in the hope it can make it in another trade. it doesn't take into account the loss in the calculation and while it show in the UI that a profit is being made, The fact is a big loss is being made.
and in the end the profit will be marginally small.

I really do understand your point of market making but at this point I'm failing to get K do the right thing.
I stopped doing manual trade since two weeks, trying to get the Bot do that for me.
doing manual trades I used to make minimum 50$ per day for 1500$ investment. with the bot I can only make ~5$ per day with hundreds of useless transactions.

If i need to tip the developer to get the right settings please let me know.
I actually decided to stop using it for now,

I'm really sorry again, but literally since two weeks and i'm trying to get K to trade right but failed miserably. at least I didn't lose the money if this point might help anyone deciding to use this bot.

Thanks again for all the help.

[Camille92]

Send me a way to contact you privately I can share some of my strategies given the size of your stack :)

[AZm1n]

Thanks Camille.

[CarlitoGrey]

@Camille92 do you have skype? Or IRC? I'm having some problems with STDev, I'm not sure it's working as I expect.

[ctubio]

ima leave this here https://kiwiirc.com/client/irc.domirc.net:6697/?theme=cli##tradingBot

[CarlitoGrey]

For example the below is 15 minutes of trading. I have STDev to 30s to be very responsive and I have changed the factor to 3 about half way through just so it gives a nice wide margin. During this time the bid is ~1euro away from the top bid. Width is set to 0.01. Is this what you would expect, I see a lot more trades on your chart.