League of Legends

We Can Pick You a Champion That Fits In the Composition With Decision Tree OVA & Matrix Factorization Models

LeagueofLegends10 - We Can Pick You a Champion That Fits In the Composition With Decision Tree OVA & Matrix Factorization Models
Loading...

TL;DR:
Lola - We Can Pick You a Champion That Fits In the Composition With Decision Tree OVA & Matrix Factorization Models


Lola.gg now has a pre-game feature that generates champion pick suggestions based on both teams' compositions, lane counters, ADC/Support partners, latest patch's meta, your champion masteries and play-style. Picks are generated using advanced machine learning models and are separated into 3 categories: Lola's picks, comfort picks, and new picks. All done automatically in real-time while you are in champ select. How it looks:
9GFyd6q - We Can Pick You a Champion That Fits In the Composition With Decision Tree OVA & Matrix Factorization ModelsChamp Select – GIF – Imgur

Lola Desktop App is in Free Open Beta!

Lola is a Riot-reviewed 3rd party app (App ID 486007 & 465359) that is downloadable from Microsoft Store for free. We recently added a pre-game feature to the app, and that is a real-time contextual pick assistant. Lola can generate personalised champion picks based on the below factors:

  1. Your team composition so far

  2. Opponent team composition so far

  3. Lane counter

  4. ADC/Support partner synergy

  5. Latest patch's meta

  6. Your personal champion pool (masteries, time played)

  7. Champion play-style similarity


How the real-time pick suggestion will look like in-app: Champ Select – GIF – Imgur

Best of all, this happens in real-time automatically while you are in champ select. No messy Ctrl+C Ctrl+V required! Lola will even categorize enemy team's picks into their respective lanes based on probability. Once it is your turn to pick, Lola will generate several picks separated into 3 main categories below. This happens instantaneously. If you hover your mouse over them, each pick will have detailed explanation/scores of Lola's reasonings. This is relative to your own champion pool, including the free champion rotation.

  1. Lola's picks: 'overall' picks that take into account of everything, a holistic view if you may

  2. Comfort picks: emphasise on your personal champion pool, but other factors will still account for the consideration

  3. New picks: discover new champions to learn (but try this in normals not ranked!)

You can then lock in one of these picks in-app without even going back to your game client! Simply click on one and click the 'Lock Champion' button. (Lock-in – GIF – Imgur)

Загрузка...

Don't just look at win rates! Consider the lot.

How does Lola decide all these factors? Lola's brain consists of, not one, but several machine learning models trained with Diamond rank and above's match data sourced over 11 regions using Riot's API. The methods used to train these models are matrix factorization and OVA based on a decision tree binary classification algorithm.

Completely optional, but if you are interested in the basic technical details:

  1. Matrix Factorization: the input data form a matrix, which will then be decomposed or reduced into its constituent parts. Coordinate descent method is used in Lola's case to find the low-rank factor matrices, in another word, the most probable recommendation. Netflix and Amazon uses something similar for their movie/product recommendation for users.

  2. Decision tree-based OVA: OVA or One-Versus-All (not to be confused with the quirk in the anime My Hero Academia) uses a binary classification algorithm, an implementation of the MART gradient boosting algorithm in Lola's case, to train an unique classifier for each class. You can think of champions as 'classes' so there will be a lot of these binary classifiers. Prediction is performed by running the relevant classifiers to get the prediction with the highest score.


What else can Lola do?

Other than contextual pick suggestions above, Lola also provides ban suggestions based on current meta (
Ban phase – GIF – Imgur) and also act as an in-game cooldown timer (see our old reddit post for the in-game feature)


Enough of this mumbo jumbo, where do I start!!?

We are very excited to apply these new tech on League of Legends and we wish to share this excitement with the reddit community. We hope that you will join us on this fantastic journey:

9PJR2CDTHKDB - We Can Pick You a Champion That Fits In the Composition With Decision Tree OVA & Matrix Factorization Models

https://www.microsoft.com/store/productId/9PJR2CDTHKDB

Lola is free to download, has no ads, and does not require other apps like Overwolf, bla bla bla.

We also have our discord link at our website footer:
lola.gg

Source: Original link


Loading...
© Post "We Can Pick You a Champion That Fits In the Composition With Decision Tree OVA & Matrix Factorization Models" for game League of Legends.


Top 10 Most Anticipated Video Games of 2020

2020 will have something to satisfy classic and modern gamers alike. To be eligible for the list, the game must be confirmed for 2020, or there should be good reason to expect its release in that year. Therefore, upcoming games with a mere announcement and no discernible release date will not be included.

Top 15 NEW Games of 2020 [FIRST HALF]

2020 has a ton to look forward to...in the video gaming world. Here are fifteen games we're looking forward to in the first half of 2020.

You Might Also Like

Leave a Reply

Your email address will not be published. Required fields are marked *