[QUOTE=Ott;49122486]It's amusing at first, but you really should get away from "lol randumb" queries. [url=http://storage.googleapis.com/books/ngrams/books/datasetsv2.html]Google Ngrams[/url] would really help in making the text seem more natural.[/QUOTE]
I concur. While Google N-Grams are cool, how exactly would I make use of them? Can you give an example of how they might be implemented?
[QUOTE=Rocket;49122631]I think he means that you should be weighting entries by their actual usage.[/QUOTE]
That's essentially what a filter does. I do have some general queries in there, though. But they're isolated to block elements, so I could just weigh them to occur less frequently.
[editline]15th November 2015[/editline]
[QUOTE=Sidneys1;49122652][img]http://i.imgur.com/0tTjQSo.png[/img]
[URL="http://pastebin.com/V0VMTpYt"]This might help[/URL][/QUOTE]
Well color me impressed. This could come in handy! Thanks!
I know that this is on a much lower level than the usual stuff here, but I started working on my game (while learning opengl and stuff. this also counts as my first project in c++ that's not a barebones console application).
So far, this is where I got:
[T]http://i.imgur.com/q1jmTdt.png[/T]
In this menu you can select shit and stuff. Start doesn't work yet, but Quit does.
There's also music playing, didn't show it in a webm or something like that since it's just a looping mp3. But it's there!
Now, since this is all using hardcoded vertices I'm working on a object loading system and a coordinate system (in general before working on more stuff I need to nicen up the program a lot).
[QUOTE=Naelstrom;49122124]I don't recommend porting to C++, just write some modules if you need C-like speed.[/QUOTE]
It's not speed I'm after, its maintainability
I wouldn't get much speed boost if I wrote it in C++
Working on a project right now that's taking bitcoin values from a JSON source, then calculating a MACD line and stochastics, and should be able to automatically trade with some decent accuracy.
Writing it in java, and it's my first major project. Working on calculating slopes within a period right now, planning on making it open source when it's done.
[editline]16th November 2015[/editline]
[t]http://i.imgur.com/m69812S.png[/t]
Pretty data. From about a week ago, after 22 + 9 elements (due to current settings) it becomes perfectly accurate
[editline]16th November 2015[/editline]
To calculate slope of a period, my current plan is to take a period in the middle that is to be examined, then look at an offset to the right and left of it and calculate those slopes as well, to verify that it isn't just a statistical outlier. Thoughts?
this weekend i went this 24 hour hackathon thing at my university, and i made this in about 15 hours, went home and passed out.
[t]https://i.imgur.com/Lot2ulI.png[/t]
its a top-down shooter that uses grid based partitioning and vector field pathfinding for enemies, it can support about 1500 of the red guys before it starts to lag really bad on my laptop. Going to try and clean up the bugs and throw it onto github at some point.
By the way, if anyone wants to help with the trading algorithm, I'm all for any help I can get.
[QUOTE=No_0ne;49123669]this weekend i went this 24 hour hackathon thing at my university, and i made this in about 15 hours, went home and passed out.
[t]https://i.imgur.com/Lot2ulI.png[/t]
its a top-down shooter that uses grid based partitioning and vector field pathfinding for enemies, it can support about 1500 of the red guys before it starts to lag really bad on my laptop. Going to try and clean up the bugs and throw it onto github at some point.[/QUOTE]
So does that mean you calculate the vector at ever grid point that goes to the player, then have all units draw an integration curve over it?
[QUOTE=DoctorSalt;49124422]So does that mean you calculate the vector at ever grid point that goes to the player, then have all units draw an integration curve over it?[/QUOTE]
I use a breadth-first search pattern to determine the distance of each tile from the player's tile, then i set each tile's movement vector to point to an adjacent tile that has the lowest distance value. Each unit knows what tile it's standing on and just looks at that tile's movement vector to know which direction to accelerate. Once the unit is within one tile it'll accelerate directly towards the player. The end result looks like this.
[t]http://i.imgur.com/UZMqsBp.png[/t]
[QUOTE=No_0ne;49124540]I use a breadth-first search pattern to determine the distance of each tile from the player's tile, then i set each tile's movement vector to point to an adjacent tile that has the lowest distance value. Each unit knows what tile it's standing on and just looks at that tile's movement vector to know which direction to accelerate. Once the unit is within one tile it'll accelerate directly towards the player. The end result looks like this.
[t]http://i.imgur.com/UZMqsBp.png[/t][/QUOTE]
How often does the path to the player update? If it isn't already, you could speed it up by making it so the path only gets updated when the player changes the tile they're on.
it does that already
[QUOTE=No_0ne;49124608]it does that already[/QUOTE]
Sorry, bad reading, thought you said when it went over 150 of the red things.
[QUOTE=Radical_ed;49123927]By the way, if anyone wants to help with the trading algorithm, I'm all for any help I can get.[/QUOTE]
Implement EMA - Exponential moving average
[QUOTE=Fourier;49124995]Implement EMA - Exponential moving average[/QUOTE]
[t]http://i.imgur.com/qHWpRdV.png[/t]
the Macd is the difference between the 12 and 26 day moving averages, and the macd signal line is the 9 day EMA of the Macd.
Oops I am blind :v:. Hope the trading strategy works :D.
[QUOTE=Fourier;49125188]Oops I am blind :v:. Hope the trading strategy works :D.[/QUOTE]
Whether it works or not, I'm just happy to have a big project. The plan right now is to detect divergence between the slopes of the MACD, and the average of the 12 and 26 day SMA's by getting the difference between their slopes within a given period, and if a divergence is detected, look at the slow stochastic oscillator to check for overbought/oversold (80/20) to finalize the decision to buy or sell.
funnily enough, I was also planning on implementing a Fourier analysis at some point to help detect patterns within the market.
I wanted to create a simpler logo for Rant 3. Came up with this quick concept. How's it look?
[t]http://i.imgur.com/Vx7LyRP.png[/t]
The version number is just there for show; it will probably only be present on the official website.
Working on a small 2D engine using SDL2, OpenGL and NanoVG
[video=youtube;xmaMMD80heo]http://www.youtube.com/watch?v=xmaMMD80heo[/video]
I'm bored, so I've created a small python countdown script that tells the countdown every time I open the terminal
[url=http://i.imgur.com/ENCjQ5G.png]
[img]http://imgur.com/ENCjQ5Gl.png[/img]
[/url]
[QUOTE=Radical_ed;49125213]Whether it works or not, I'm just happy to have a big project. The plan right now is to detect divergence between the slopes of the MACD, and the average of the 12 and 26 day SMA's by getting the difference between their slopes within a given period, and if a divergence is detected, look at the slow stochastic oscillator to check for overbought/oversold (80/20) to finalize the decision to buy or sell.
funnily enough, I was also planning on implementing a Fourier analysis at some point to help detect patterns within the market.[/QUOTE]
Fourier analysis might be useless here I think.
Do auto-correlation instead, you will detect patterns much better here.
[QUOTE=No_0ne;49123669]this weekend i went this 24 hour hackathon thing at my university, and i made this in about 15 hours, went home and passed out.
[t]https://i.imgur.com/Lot2ulI.png[/t]
its a top-down shooter that uses grid based partitioning and vector field pathfinding for enemies, it can support about 1500 of the red guys before it starts to lag really bad on my laptop. Going to try and clean up the bugs and throw it onto github at some point.[/QUOTE]
There was a hackathon at my university this weekend as well; Are you in the western New York area? I was supposed to go but decided to stay in my room and work on my project alone, since I don't know anybody here yet :(
I'm researching web traffic patterns as an undergrad!
Holy shit. The professor in charge took one look at my resume and said, "Well, I could give you the same old typical interview questions, but you look well beyond on that. Here are these research papers. Go read them. "
Life is good. So who has some good literature/experience on the subject? This is my first time doing official research as an undergrad, what should I know?
[QUOTE=Winner;49125547]how well does it scale down?[/QUOTE]
128x128
[img]http://i.imgur.com/R2Ed3Xr.png[/img]
64x64
[img]http://i.imgur.com/owiibWk.png[/img]
32x32
[img]http://i.imgur.com/i7fLYt5.png[/img]
[QUOTE=Fourier;49125718]Fourier analysis might be useless here I think.[/QUOTE]
You would say that.
[QUOTE=Fourier;49125718]Fourier analysis might be useless here I think.
Do auto-correlation instead, you will detect patterns much better here.[/QUOTE]
It's not particularly useful for patterns but it'd give a nice idea of how erratic it is.
[QUOTE=Berkin;49126912]128x128
[img]http://i.imgur.com/R2Ed3Xr.png[/img]
64x64
[img]http://i.imgur.com/owiibWk.png[/img]
32x32
[img]http://i.imgur.com/i7fLYt5.png[/img][/QUOTE]
The last one needs either hinting or an outright alternative version.
It looks really blurry as is.
[QUOTE=chaz13;49127321]It's not particularly useful for patterns but it'd give a nice idea of how erratic it is.[/QUOTE]
Using it to determine how large the trade period should be (high volatility = shorter timeframe, less long-term securities) interesting, at least.
Best part is that the entire blockchain is downloadable, so even if an idea is totally out there and might not function, I can test it on years and years of old data for the hell of it to verify.
[t]http://s13.postimg.org/ezbwa1k93/muru_2015_11_16_19_11_41_69.png[/t]
Alright so.. I have now added fog of war and level generation. Though the level generation is really really simple for now due testing purposes.
I would love if a few of you could test it out and tell me how it runs on your pc, if the fps is stable (shown in the upper left corner), if the generation takes too long and what your system specs are (especially ram).
Controls are pretty easy: W, A, S, D for moving, left click for removing blocks, and esc for ending the game.
[url=https://www.dropbox.com/s/ved3ti1utud495u/test_fps.zip?dl=0]Download[/url] (Just unpack and run muru.exe)
[QUOTE=Da'Jobat;49127190]You would say that.[/QUOTE]
FFT would be insanely useful for speeding up correlation process, but Fourier coefficients wouldn't give you much of idea. Ok well maybe maybe, but you'd need to put weight, so older data - less weight.
Test it anyway, maybe I am wrong very much.
Still autocorrelation would give you much better idea of "periodicity".
Sorry, you need to Log In to post a reply to this thread.
