Predicting price movements

[Archery1969]

Interesting read at https://bura.brunel.ac.uk/bitstream/243 ... llText.pdf ?

I havent got a clue what they talking about.

Anyone ever tried to implement it ?

[sniffer66]

Interesting read at https://bura.brunel.ac.uk/bitstream/243 ... llText.pdf ?

I havent got a clue what they talking about.

Anyone ever tried to implement it ?

Intersesting read. I've read a very similar paper. probably by the same author, for predicting the outome of tennis matches using machine learning. Also with similar results

The technical skills are way beyond me at the moment though

[sniffer66]

Found this an interesting read as well, and much less complex

https://medium.com/swlh/predicting-atp- ... e03d99f410

I've been attempting to use the findings in my tennis bots for a few weeks now

[goat68]

Interesting read at https://bura.brunel.ac.uk/bitstream/243 ... llText.pdf ?

I havent got a clue what they talking about.

Anyone ever tried to implement it ?

It's sort of what I do, and probably a lot of algo types, ie.test strategies with backtest data, and then use AI (my brain for me!) to tailor/improve. Interesting if you look at the results section you can see the issue most people hit, ie.you train for 3 months of back data, then forward test and it all breaks down!
Also I think from my experience the "Placing Orders" paragraph is a bit lacking, their simulation for order matching/filling is limited, I see no mention of estimating fill rate/cancelled orders on queues, maybe that is in ref [25].
The document also implies they have not "live tested", only forward tested with the emulator, the last sections refers to this as something for the future. My biggest hurdle was ensuring my "simulation" is as close as possible to "live testing", and it only takes a small difference to have large deviations.

[LinusP]

Screenshot 2021-04-20 at 09.21.13.png

Anyone else struggle to understand how the ROI has been calculated or is just me? Assuming £2 bets I would assume its more like 4%.

Not a single mention of EV

[Archery1969]

Screenshot 2021-04-20 at 09.21.13.png

Anyone else struggle to understand how the ROI has been calculated or is just me? Assuming £2 bets I would assume its more like 4%.

Not a single mention of EV

I struggled with the entire document.

[sniffer66]

Screenshot 2021-04-20 at 09.21.13.png

Anyone else struggle to understand how the ROI has been calculated or is just me? Assuming £2 bets I would assume its more like 4%.

Not a single mention of EV

Good point. I just looked at the high level numbers. It's way off

[ShaunWhite]

With papers like this it's best to read the conclusion first, and there were so many parts that needed red pen I didn't bother reading the rest. If they can conclude algo trading is high risk/high return speculative trading then they've missed the whole point. And what's with the crappy 1400 samples.

There's no shortage of papers like this knocking around, only a few ever get some sort of peer review
http://www.mathsportinternational.com/ ... stract.pdf

[Euler]

Found this an interesting read as well, and much less complex

https://medium.com/swlh/predicting-atp- ... e03d99f410

I've been attempting to use the findings in my tennis bots for a few weeks now

The problem with that article is that it's sort of got the wrong end of the handle hasn't it?

First serve % is directly linked to second serve so there is no need to weight it, the metric exists already.

I'll check my assumption of the article, but that would be me immediate view.

[Euler]

Interesting read at https://bura.brunel.ac.uk/bitstream/243 ... llText.pdf ?

I havent got a clue what they talking about.

Anyone ever tried to implement it ?

I always think the funny thing about a lot of academic papers is how they completely miss the point. It's a sort of intelligence trap.

I've only ever done one presentation to a math society who were interested in what I was doing and after 1/2 an hour the penny dropped.

I'm always surprised at how obtuse people get in solving fundamentally simple problems.

[sniffer66]

Found this an interesting read as well, and much less complex

https://medium.com/swlh/predicting-atp- ... e03d99f410

I've been attempting to use the findings in my tennis bots for a few weeks now

The problem with that article is that it's sort of got the wrong end of the handle hasn't it?

First serve % is directly linked to second serve so there is no need to weight it, the metric exists already.

I'll check my assumption of the article, but that would be me immediate view.

Are you basing that conclusion on percentage in, rather than points won on 1st\2nd serve ?

[Morbius]

Interesting read at https://bura.brunel.ac.uk/bitstream/243 ... llText.pdf ?

I havent got a clue what they talking about.

Anyone ever tried to implement it ?

Advanced AMMs (automated market makers) already use variations on this or at least components of it.