Greedy Goblin

Wednesday, October 5, 2016

Lucidity and statistics

Reading guides of good players is an accepted norm of learning to play. Following the advises of successful people is considered a good practice in life. For some weird reason people are still bad at games and unsuccessful in life. Maybe there is something wrong with those guides?

No and yes. The personal success of the author should give support to his guide. Unless purposefully misleading, he wrote down things that helped him. But yes, they aren't useful, because both top players and successful businessmen are in the top 1% and you can't really get there without something special: extraordinary dedication, unique talent, luck or connections. There cannot be guides that lead you to the top 1%. Don't get me wrong, no dedication, talent, luck or connections will get you anywhere if you do stupid stuff and them following their guide was key to their success. But it won't help you much.

What you'd need is a "how to get to the top 10%" guide". The top 10% is big enough to welcome everyone without connections, natural born talent, luck or no-life, needing only decent work and smart (but not genius) decision making. But such guides are usually not written, because people getting into the top 10% don't see their feat worthy enough to publicize and will also be talked down by top 1% people.

The net is full of Warwick guides and I boldly went against them, knowing the above. But "go with your own decision" is probably not the best idea as you aren't all-knowing. But now I've found the source of the true "how to be above-average" guide: statistical analysis of thousands of games with Warwick. Click on the items tab and short them by winrate and see how players with that item fared in ordinary (not top 1%) games. For example my favorite boots, Lucidity has 59% winrate, while the most popular pick has only 53%. I used it for its cooldown reduction so I don't need to have it from other items. Another item standing out of the crowd is the Guardian Angel that can make or break a teamfight at late game. It allows you to charge in head-first without using your ultimate, die eating all the cooldowns of enemies, watch the fight during the resurrection picking an enemy, use your ult on him to score an instant kill and refill your HP. Hint: if you return to the shop with Angel being on long cooldown, just sell it and buy something else instead. Buy a new one 5 mins later.

Thornmail is another great item, allowing rounding up a huge minion wave in no time, faster jungle pack clearing and a nasty surprise in PvP. You can find all the good items in the 55-57 bracket, pick as you please. The starter item section shows that my pick (machete and health potion) has the top winrate too. In the overview section you can see that Ghost/Smite beats Flash/Smite, something I will switch to. The masteries section show that I was right going to the defense tree instead of the usual Fervor of Battle, though my masteries also need tweaking (used Bond of Stone). I stick to 6/6/18 instead of the suggested 0/12/18.

However the greatest surprise is that red enchantment: challenging smite - bloodrazor, which is considered "must" by every guide I've seen is beaten by both cinderhulk and chilling smite - bloordazor. I use ward -bloodrazor by the way which is so rare that it's not even listed and I won't ever get rid of it. Those extra wards saved me countless times.

Anyway, trust the statistics that tell the results of real people around you, instead of guides of the top 1%!


Vizjira said...

While i do prefere over guides everything you see there has to be seen in context.

For example our beloved Hypercarry Vayne has a 66% winrate with Bloodthirster. You might think getting this as soon as possible will help you win games but what it actually means is if vayne gets to the lategame (and BT is core of her lategame build usually last or second to last choice) she has the highest winrate of all adcs due to beeing vayne.

Secondly those champ stats are gathered from Plat+ players which is top 10% therefore potentially more mechanical demanding than lower tier builds and players dont distribute evenly across all builds - experienced players make different item choices, but they might not work in every context, while novice players follow guides/meta etc more strictly.

Lucid boots has 280 plays on eune, that can lit. be a influenced by single player/smurf grinding out ladder with his fav champ and build. Compare that to EU 430 plays at 50% winrate or Korea 530 at 53% (in both cases not the best item). Thornmail is not played at all on most servers and has a small sample size (with good w/l) on those it has been played. is nice to figure out in which direction a champ can develope if you are thrown into the cold water with him, but you really want to go, look up dia+ mains of your champ and figure out what people do with those items, also beeing able to watch a game/teamfight over and over is invaluable.

Anonymous said...

Worth noting is that the expectations you can have towards your teammates change drastically as you rise through the ranks.

In Silver/Gold, buying Ward smite on any jungler is a very solid choice, as few people understand the concept of warding. Once you get to higher ranks and encounter more competent teams, buying and placing wards (pinks) is something expected for every role, providing the team with more vision and letting you maximize your own impact (red/blue smite).

Champion guides are typically written by players in higher tiers, who live in a vastly different play ecosystem from you. Their advice might be optimal once you hit Plat/Dia, but it might not be in Silver/Gold where you must compensate for basic shortcomings in your team first (lack of warding, horrible map awareness, etc.)

Fidtz said...

Cracking post, back to PuG/blue-raid era Gevlon type things.

I would only add to beware of low pick talents with high win rates. They may only be picked for fun when you would bascially be able to win with anything due to team imbalance or to hard counter certain other picks when without that counter to do they are useless.

Anonymous said...

Those statistics are biased.

-Some items are only good to build if you're already far ahead. However, these items are far more likely to correlate with a win, because if you're already far ahead, you're more likely to win.

-Some items are only good in certain matchups. For example, you can see how thornmail is only good against a team that mostly uses autoattacks/AD. It's possible that Warwick happens to do really well against teams that mostly use AD. If you want a clear example, look at Rammus, whose winrate is much higher when he builds the first 5 items specifically to protect against AD champions. That wouldn't work at all against a team of all AP, but nobody picks Rammus into heavy-AP teams.

-If the statistics include bronze/silver games, these will be the lion's share of the item builds, and the statistics will not be useful above gold. If the statistics are something like (Platinum+), then they're already talking about the top 1%, more or less. But the reason for this is that people in bronze, silver, (gold) aren't good enough to follow simple build paths, or want to try something 'fun'. All in all, you should never look at bronze-gold build paths anyway; if they are successful, it's only by following the plat+ players.

Samus said...

A couple of notes on something like this.

It is generally accepted that you should build offense if you are winning, defense if you are losing. The result is that in every statistical analysis like this, the offensive items and builds will have higher winrates than defensive ones. For instance Bloodrazor-->BotRK-->Wit's End has a 62.86% winrate, but you would only build that if you were 5-0 before 20 minutes, already very likely to win. I assume you will be switching over to Dead Man's Plate-->Spirit Visage, the highest winrate core build and all the more notable that you would build it even if you weren't ahead.

You should also look at how often something is built at all. Many items, particularly like Guardian's Angel, are typically only built as 5th or 6th item. Usually, only the winning team will get that far in their build, the losing team wouldn't get past 3rd or 4th item. Guardian's Angel is still a solid item, but there's a reason you don't see it in any of the highest winrate "core" builds, despite a 67% winrate on its own.

You should also be aware of situational items, which will usually be built only in certain situations, but are great counters so will have high winrates in those situations. Merc Treads is the prime example of that, something you would build if the enemy has lots of CC. Thornmail also fits that description and is not as frequently built. If players are only building Thornmail when the enemy has lots of AD auto-attacking champions, it is a fantastic counter and will have a very high winrate.

For you in particular, the BotRK has a high wirate because Warwick can make use of the item's active ability to gank with (slows the enemy and gives you a speed boost). If you think you can master that, BotRK might be a better choice, but if not stick with Wit's End.

As far as the warding jungle item that no one builds, I noticed that in your two games that you posted none of the 4 supports built a Sightstone. That is basically unheard of by mid to high Silver, you will get flamed hard by your team if you don't build one as support as one of your first items. Even with the Sightstone and yellow trinket, lower level players are terrible at laying wards. So currently, I am guessing you are supplying most of your team's vision. By high Silver, your team will be doing MUCH more for vision and the warding jungle item will not be necessary.

One final note:

"OP.GG uses data from ranked games Platinum and higher."

This already IS data from the top 1%. You will want for data from all ranks. It will also be a MUCH larger data set, many of the winrates on OP.GG will be off from small sample size.

Slawomir Chmielewski said...

If you are in the bottom 30% don't do what the man in top 1% is doing. Do what he did when he was in bottom 30% himself. If he started from higher level (by birth or whatever other reason) ignore his advice entirely - it isn't applicable.

Game_Analyzer_MK3 said...

Another thing people should consider about the top 1% is that the top League player is probably also showing off and grandstanding for the camera, and would never do some of the stunts he pulls in a "real" match where he was going for ELO. They have too high a chance of failing and causing a loss. Many League players are being entertainers, not doing real plays that even a Diamond player would do in a "real" match. After he is done entertaining, he can always grind the ELO back, and cut-out the worst failures from his video. Just like TV, you only see the "final cut" of these Diamond player videos. If a regular player tries to pull the same stunts, then the results will be even worse for the regular player.

Replays have a lot of problems, but playing matches in League is tiring. People have only so much energy in a day. I've noticed that a lot of players get tired and start doing really silly things around the 40 minute mark in matches. I wonder if a five minute break between matches would significantly increase win rate. You do have to play games to get better, but watching replays is a lot less tiring than playing games and provides a different perspective. Still, watching replays of Diamond players doing stunts for the camera in perfect match-ups has many problems. You don't have to watch Diamond players though. Real high silver/low gold players playing against each other in a mostly normal match are available. League of Graphs has replays of such games available using LoL's own client. Watching other people's matches is limited, but it does provide real information in a less tiring way and with a fast-forward button. Watching your own matches in replay can also be helpful, you can see things that you missed while you were distracted by having to play. Finally, most players in a region seem to follow the same "meta", and have difficulty dealing with any change from it, so your region is probably the best to watch so you learn it's "meta".

Raphael said...

This is incredibly applicable to World of Warcraft and doing top DPS. I am working on a blog post about it, and how looking at other people's logs is one of the best ways to improve your DPS; I can let you know when it's done if you like.

For anyone who plays trading card games like Magic the Gathering: "Guides" are like reading an article before any big tournaments in a new expansion, with good players speculating on what will be good. "Looking at Logs/Statistics" is like netdecking - you know exactly what worked, and use it yourself. What's better, speculation, or knowing exactly what worked?

What's even better for World of Warcraft PvE is that your opposition doesn't change. Xavius doesn't change his reaction if you use a better DPS rotation. In PvP games, over time, the metagame can shift. In PvE, unless developer introduces a patch (usually announced beforehand!), better strategy is objective and gets better results.

In League, with such a large player-base, it takes a long time for the meta to shift especially at a lower skill bracket. Changing your play to keep up with what has been proven to work is likely to work well for a long time.