HNHacker News
TopNewBestAskShowJobs

noxvilleza

317 karma · joined November 11, 2016

E: bensteenhuisen [@] {google's mail site} . com

Technical Director @ Team Liquid. Live in Berlin, sometimes around Utrecht! I'm interested in statistics in competitive video gaming, mostly Dota 2; as well as mathematics, film, and trivia. I run datdota.com and windrun.io

submissionscomments
noxvilleza··on Implementing the Elo Rating System (2020)
Depends on how quickly you calibrate to your 'true rating'. Elo-like systems take relatively quite long, since you can gain at most k points per game (but obviously fewer points per game if you're being matched with someone who's similar in skill to you). Chess GMs for example stream their 'new account to GM' runs, and it takes very many hours.

Chess is a game where there's very high correlation with individual skill and the outcome, in more noisy/random games like Dota 2 for example (which also are ~40 minutes on average) your expected relative gain per match is much lower.

noxvilleza··on Implementing the Elo Rating System (2020)
Although highly dependent on the game itself, I did some investigation back in 2008 into testing various Elo modifications on a pretty large dataset of individual players when playing in a 5v5 team game, and geometric mean of individual ratings was the best simple predictor for which team would win (the update method was essentially 10 individual matchups of each player vs the aggregate rating of the enemy team).
noxvilleza··on Implementing the Elo Rating System (2020)
Glicko (1 and 2) don't handle complex matchups well (as in, anything that's not 1v1). Trueskill was specifically designed to extend the Bayesian concept of Glicko in a more general way which allowed this type of use-case.
noxvilleza··on Implementing the Elo Rating System (2020)
Elo-like systems (or various other rating systems) are often as part of a matchmaking algorithm - since you can easily calculate expected game win percentages for each party, and this allows you to create more enjoyable matchups. An example of this happens even in online chess: you're much more likely to be paired against a player with a similar rating, because you're much more likely to enjoy a match against someone similarly skilled.
noxvilleza··on Google could have killed Facebook with the flick of a switch
Sounds like the South African government approach, except they released their own browser with the last Flash version: https://www.zdnet.com/article/south-african-government-relea...
noxvilleza··on Databricks is an RDBMS
... and if it doesn't fit on Excel on a laptop, the natural path is just some database running on a desktop, then a server, then a beefier server. It's actually amazing how far just something like postgres will scale for your data needs.
noxvilleza··on How to Turn an IKEA Coffee Table into a DIY Server Rack
Not a full server-rack but had 4 desktops (2-side by side on either side of my small desk) running a variety of stuff in my first two years of university (CS1.6 servers, WoW emu server, network bridge, teamspeak, DC server) - one was my personal computer. In 3rd year it went up to 5 since my room was slightly bigger and my friend would often come play games in my room (he didn't have internet).
noxvilleza··on How to Turn an IKEA Coffee Table into a DIY Server Rack
I use my Lack to keep my desktop off the ground but also not be on my table. It's a reasonable height next to my desk, and there's space for a multiplug. The value!
noxvilleza··on The Future of RTS
Definitely not directly comparable - since every civ has access to some common/baseline technology, but certainly differentiation (peaks/dips) in specific areas. That said, SC2's races are more different but not enough IMO to say the game is overall way more diverse.

Why do you say 'The downside is that it makes balancing nearly impossible.' when referring to MOBAs?

noxvilleza··on Lawyers automate this, so why don't airlines?
Slightly different but similarly annoying situation happened to me. Was flying back from Birmingham to Berlin, via Dusseldorf (with Eurowings, booked via Lufthansa). There was bad weather at Dusseldorf so the outgoing flight was delayed by an hour or so. Knowing that any passenger arriving at Dusseldorf would arrive at ~8:30pm, and thus miss their connecting flights for the day - they still insisted on people flying there. We arrived, and were stranded - obviously. Their agents closed the booth and tried to leave, but after a bunch of discussion they then wanted to put people on buses or trains to various parts of Germany overnight. A few of us demanded accommodation and ended up staying in the nearby hotel. After like 10 emails back and forth over the next week with the airline they ended up refusing to pay compensation, so I raised a claim with the consumer arbitration board in the country I live (soep-online.de). Two days later I get an email from the airline saying they'll pay out. It seems so utterly scummy that they try lie and deceive consumers out of their rights - how are there not strict punishments in situations where airlines mislead and screw over consumers?
noxvilleza··on The Future of RTS
Whilst I certainly have huge respect for HuK, I feel that either his use of the categories in the article as perhaps a bit simplistic. Many of the points he mentions are really intertwined - most importantly balance vs variety: it's very hard to score highly on both.

SC2 has 3 very distinct races (he gives it a 8.8 for variety), hence there are just 6 possible matchups. AoE 2 (DE) has 35 (he gives it a 3), making over 600 possible 1v1 matchups. This creates a huge amount of extra work to ensure matchups are enjoyable - yet (at 1650+ Elo) there are just 2 races with > 55% winrate (Franks @ 56.98% & Celts @ 54.21% on random map). Despite the good balancing, tournaments also come up with rules to limit the impact of the best civs, normally with specific civ bans and limited picks in a series. It's really only because of the huge overlapping tech & unit tree that this is possible - if there were more diversity there would be too many variables to balance around. This is further compounded by the variety of maps which the game is played on - as he does mention the map does affect which of these are viable, and this makes the game even more interesting and diverse since you need to adjust to multiple types of map. SC2 on the other hand has a much smaller responsibility in this regard: provided there are interesting viable builds within each race the task to balance across-races is minimal. SC2 also only has 7 maps in the pool (at least, right now) and just spawn positions are randomized. This makes preparation way more important, and less dynamic.

Both AoE and SC2 also use the space as a stop-gap for specific strategies - it's a way for the less aggressive player to fend off early aggression, but also a way for a more dominant player to choke out resources. Traditionally this is done by controlling the entry to your main base or your expansion (in SC2), but in AoE it's way more open and difficult to accomplish this most of the time. This importantly makes players rely on very adhoc dynamic walling, often incorporating undestroyable natural resources (like gold deposits).

noxvilleza··on Curry Triangle
When I saw this thread this is exactly what I was expecting (read: craving).
noxvilleza··on UK and EU agree Brexit trade deal
All over this thread you seem to mention potential trade benefits for the UK based on being separate from the EU. Can you name some benefits (perhaps let's say 3) that you expect will be negotiated more favourably for the UK within a year that the UK was unable to get as part of the EU.

Also ... being part of a a trade bloc always has benefits and positives, but the two key negatives touted as 'deal-breakers' in the Brexit negotiations have been related to (1) sovereignty and (2) concessions. A few weeks ago Peter Verovšek wrote a short but pretty good article on effectively the illusion of sovereignty (https://www.socialeurope.eu/brexit-and-the-misunderstanding-...) for the UK, and in terms of concessions it seems that a small nation with limited resources hardly makes a net loss by being part of a larger trade bloc: access to a huge market while exposing a small market of your own.

noxvilleza··on The FBI Is Breaking into Encrypted Devices – We’re Suing
This situation is a bit more complex because it's contested (up to 11th circuit court of appeals) that "the decryption and production would be tantamount to testimony by [the suspect] of his knowledge of the existence and location of potentially incriminating files; of his possession, control, and access to the encrypted portions of the drives; and of his capability to decrypt the files." - and if it's testimony it's covered under 5th Amendment protections.

Quite interesting 4-page summary of the topic by the Congressional Research Service from earlier this year: https://crsreports.congress.gov/search/#/?termsToSearch=LSB1...

noxvilleza··on Chess tactics explained
Yeah overall some theme diversity. Would also be interested to know the median Elo rating by tag - since that'd serve a good sort of table-of-contents of order for people to learn chess tactics.

What is also pretty interesting would be a (fuzzed) per-tag Elo rating on the user - so they can see which themes you are good/bad at relatively (for example, if you constantly get problems on Trapped Pieces wrong your rating on those problems would be lower than your average Puzzle Rating).

noxvilleza··on Chess tactics explained
A bunch of video games do this ('shadow banning / shadow pooling') - with varying degrees of success. A lot of the cheaters still just create new accounts every so often.
noxvilleza··on Chess tactics explained
Really love your chess puzzles, I play Puzzle Rush most days! If you can share, how many puzzles do you add each month? Are they mostly hand-generated or do you have some nifty way of deriving puzzles from real games automagically? Would also be really cool if a puzzle rush spread out puzzle types more it feels like there is heavy clustering of a specific problem type at key rating ranges: early puzzles (1-6) are very commonly back-rank checkmates; 9-12 are discovered check, etc - it would be nicer if the problem meta-tags were not re-used a lot within a single run.
noxvilleza··on Chess tactics explained
An easy way to detect cheaters is to look at overall accuracy as some aggregate metric of deviation from 'optimal' (as per the engine(s) used). This could just be a average of squares of centipawns lost per move, or classifying moves like how chess.com does (brilliant/best/excellent/good/inaccuracy/mistake/blunder/missed-win) and then looking at average distributions from this.
noxvilleza··on Chess tactics explained
It's "Elo" (not an acronym). There is the Computer Chess Rating List (https://computerchess.org.uk/ccrl/4040/rating_list_all.html) which has the current rating, on some predefined hardware.

The easiest way to do what you're discussing is just giving less evaluation time to the engine - slowly increasing it. If you wanted it to be even better you could calibrate multiple engines to make an even smoother (but changing underlying evaluation).

noxvilleza··on Chess tactics explained
A lot of cheaters are detected and banned, but like you say there's probably a lot of 'minor cheaters' not caught.

The issue really is the group who are detected and banned - they just sign up with a new account, and then carry on crushing players until they're banned again. The average game quality in this situation is pretty poor - because the rating of a player is very far from their true skill.

noxvilleza··on Prediction Markets Beat Polls
In addition to just using Brier scores (or alternatives) to evaluate their performance, I'm also interested in seeing how 538's underlying Monte Carlo models shift based on x% of error/uncertainty? (for example instead of evaluating a polling advantage as a single value K within a single simulation rather consider it a normally distributed variable).
noxvilleza··on Impact of Go AI on the professional Go world
It's "Elo", not "ELO".

Afaik they analyze the deviation (in centipawns) of each played move to the optimal move calculated by popular chess engines (like Stockfish, Houdini, Komodo, etc); and aggregate in a sort of weighted Brier score.

If a player is constantly making optimal moves, they're likely cheating (they can analyze top level professional matches for upper skill benchmarks). There is some lower weighting for opening moves (which is where the database part comes in just knowing when a game moves from theoretical 'book' openings into original territory).

Each of the sites have different implementations of this: chess24 for example even partially gives your accuracy of book moves and then differentiates between Blunder / Mistake / Good / Excellent / Best move based on the deviation from optimal; but it doesn't show you how it's aggregated into a single value.

noxvilleza··on Show HN: We used 890M chess games to make an interactive opening graph
Also it's "Elo", not "ELO" (it's not an acronym).
noxvilleza··on Valve has mutated from a game developer into a financial middleman
'$179,702,427 have been given out in prize money (mostly from Valve) thus far'

Come now, this part of your response is totally misleading. Valve contribute a base $1.6M for TIs, $3M for the old majors, $150k for the current minors and $500k for the current majors. The vast majority of the $179M is crowd-funded money: money which Valve facilitated the collection of but which many people only gave to boost the prize pool of The Internationals.

noxvilleza··on OpenAI Five at Dota 2 – The International [video]
That or there's something like 200ms "windows" in which API sync occurs - so if someone Blinks next to you you can react within 1 tick (33ms) if the timing is right.
noxvilleza··on OpenAI Five at Dota 2 – The International [video]
*Elo, not ELO :)
noxvilleza··on OpenAI Five at Dota 2 – The International [video]
> The current hero pool isn't, by any means, random.

Some of the restrictions they've discussed - like initially they couldn't handle with illusions, summons or invisibility at all. They obviously dropped the invis restriction.

My feeling is that the AI just learned that of the given pool, 'deathball' was the best (and specific heroes like Gyrocopter/Sniper were the most effective carries).

noxvilleza··on OpenAI Five at Dota 2 – The International [video]
Net Worth Difference is the best single indicator, but there are additional factors like tower difference, item purchasing decisions, hero composition, hero aliveness & buyback status, aegis, etc. Some specific heroes also wildly differ from the norm for each one of these factors. The game is very complex :()
noxvilleza··on OpenAI Five at Dota 2 – The International [video]
It's actually internationally delayed reactions to make it more fair (200ms of delay added).
noxvilleza··on OpenAI Five at Dota 2 – The International [video]
They're still the ~19th best team in the world.
← PreviousPage 6 of 6