Running Gekko as a taker

[cubit]

One of the web pages I've read somewhere today said something like "I hope we can run gekko as a taker soon", and as I've already done this (with disastrous results!) I thought I would share my experience here.

I am finding that whilst the papertrader buys and sells immediately at the market price the tradebot doesn't find it so easy!

So I thought I would give it a try as a taker rather than a maker.

I found the kraken API documentation and modified the kraken.js file, line 303, from -

   ordertype: 'limit',


to

   ordertype: 'market',

I kicked off gekko, kicked off a tradebot and walked away so I missed the carnage!

I lost 20% in 12 hours because my tradebot decided to buy high and sell low.

I guess whilst gekko is quoted a market price what you actually get when you a place a market order is the closest limit order somebody else has placed and not the actual market price that was quoted to gekko!

So... question for Mike I s'pose... in theory was that all I needed and my experiment ended as it should? Or did I miss something and maybe running as a taker could produce better results than I saw?

Also I had decided that running a machine on the ground probably wasn't any worse than running a machine in the sky (I have good Internet)... being that the standard limit orders take a while to go through... but having read the Agent X arbitrage story I am also wondering whether getting as close to the source as possible would help running as a taker?

Just interested in people's thoughts on this... particularly those who actually know what they are doing as opposed to me!!

[askmike]

So... question for Mike I s'pose... in theory was that all I needed and my experiment ended as it should? Or did I miss something and maybe running as a taker could produce better results than I saw?

Gekko is using limit orders now which is considered conservative. If the is no major price move happening this means you pay less for 3 reasons:

- maker fees instead of taker fees (depends on the exchange)
- crossing the spread
- slipping into the orderbook

I have given some examples about the them here: https://github.com/askmike/gekko/issues/...-408744682

These last two are relative and depend 100% on the market you are trading on. If you are a tiny market (like most altcoin markets on Kraken, you might end up losing 1% of you balance every time you trade). That's the reason why I'm slow in implementing it in Gekko. Doing market orders is easy, but making sure we don't do it when there is no liquidity is very harder. With the current defaults it's "harder" to lose money quickly due to lack of liquidity.

[cubit]

Thanks for the response, and in particular the link which confirms my suspicions.

If that hadn't done it, it pointed me toward the backtesting page where I saw the graph (https://cloud.githubusercontent.com/asse...7433ea.png) that a couple of weeks ago I thought was going to make me a billlionare!!

I'm sure you were there long ago!

Anyway just one thought regarding the linked discussion, where you say...

"The spread is the difference between the price of the current best buyer and best seller. And when Gekko is simulating trades half of the time it's simulating buys and half of the time sells. But unfortunately it does NOT have these prices since Gekko has no way of getting them for any point in the past."
...
"Same as the spread, Gekko does not have this data for any point in time since Binance does not provide it (they only provide it in realtime)"

Could you not gather up the data in real-time and stash that centrally so that it could be used for backtesting purposes, perhaps as part of the Gekko Plus service?

You've probably already had that thought as well!!

Thanks again.

[askmike]

Could you not gather up the data in real-time and stash that centrally so that it could be used for backtesting purposes, perhaps as part of the Gekko Plus service?

Yes that's on the roadmap. But I'm definitely more up for open source first. So I don't want to offer a paid version which is a lot better than the open source version since I have all the data needed to run proper simulations.

Also note that in markets without decent liquidity it's very hard to make money with a Gekko like system (unless it only trades long term, eg. less than once every few days).