5 platforms which will increase your Dota skill. GOSU.AI, MoreMMR, Feedless, PVGNA and Gameleap under the scope!

18 min readJun 26, 2018

Hey! I’ve wasted all working day to compile and translate two articles from Cybersport.ru, and now we have one mega-long read about 5 most popular Dota 2 educational platforms. Maybe you will find them useful and those platforms will really help you to increase your skill. Enjoy!

GOSU.AI

GOSU.AI is a match analysis platform. At the moment of the first revision, a website was at the stage of alpha-testing, right now it is already beta. Portal is in English. However, it is known that part of the developers’ team speaks Russian

The system analyzes your performance in particular match, measures several variables and provides information on mistakes.

There are 8 variables totally. Sections for farming, micro control, Hand of Midas and efficiency of hotkeys remain the same. The analysis of items at the beginning of the match has been removed, and instead of “mechanical mistakes” and “team communication” analysis, which were promised, developers added the analysis of ganks, warding, damage priority and a section for cheaters detection in player’s matches.

The system now evaluates how well a player did in each section. Evaluations are equal to Dota’s rank, however, they do not seem like connected to them directly: the difference between Divine and Ancient is not necessary a few more creeps latshitted/denied on the lane. More likely it is something closer to a school grade.

An analysis No{o}one’s performance in a match against Vici Gaming. Two following screenshots are from the same game

The service has the monetization impemented. Each player has 10 daily points, and are spent one point per section in a match. In fact, you can get one full and one partially analyzed game for free. This “part” depends on if the hero bought wards and Hand of Midas.

There are two options, which could be purchased only for real money. Those functions are personal match analysis from a Divine 5 player, and highlights. There are pre-packed offers: for 5 or 20 dollars you can get more points for analysis, 125 or 500 gold coins, and an access to the Beta version of GOSU.AI for PUBG as a bonus. Gold coins could also be purchased separately, 50 coins for 2$.

GOSU.AI now also provides information on fights with an enemy.

Where exactly it happened, who participated in it, did a player managed to kill an opponent, what helped an enemy to survive and how much gold and EXP this kill was worth.

The function itself does not seem exceptionally educational. Yes, it can give you a general understanding on why you died and why enemies were stronger. However, usually, a player would understand that, for example, it was not a great idea to dive under the opponent’s tower. For better evaluation it would be great to have advice on how could a player could avoid dying there: “Did not use BKB”, “Did not use Magic Wand” or any other defensive abilities/items. Of course, not any situation is possible to survive using those skills/items. However, the most useful information here is when a player has seen enemies on the ward but did not react. This should draw the attention of a player towards minimap and show how important the information on it is.

However, Dota is a very various game, to assume that fighting under the enemy’s tower is already an exact reason of player’s death. Sometimes a fight in such a location may be the correct move, but the death was caused by the spell missed spell of a player or his teammate, and there is no such option on GOSU.AI. That’s why this feature seems mostly valuable as a map for convenient replay overview. You check out which moments are the most interesting to you, quickly check them out and personally evaluate your mistakes.

Warding is a successful part of GOSU.AI After the match, a player can check how many wards he placed, where, when, for how long they lasted, how many heroes were killed with this vision and the “quality” of every ward. The last characteristic depends on how long the ward lasted, how many kills were committed under the vision of the ward, was it placed before the player’s death and was it placed under Smoke of Deceit

GOSU.AI also learned to analyze how good a player prioritizes damage in team fights. For that, a service calculates the index, which is called “Damage/max HP”, and compares it with the developer’s presentation of damage priority between the heroes. The function works in a strange way. In fact, those recommendations and priorities are reduced to “carry and midlaners should hit opposite carries and midlaners”, while the text sometimes contradicts the picture, on which the priorities are indicated. Even for many support heroes, the recommendation is averagely the same. However, at least for heroes like Oracle, for whom the most important thing is not damage, but saving the allies, the authors recommend what is needed.

It seems like the system does not adjust to specific heroes, but simply advises that almost everyone should hit the core heroes. Such a straightforward approach does not work in such complicated game as Dota. For example, I was playing Queen of Pain, my enemies picked Io. It is obvious, that such character with high mobility and lots of magical damage have to focus Io at first, otherwise he would just save his teammates. But no, GOSU.AI is suggesting focusing Troll Warlord and Drow Ranger.

The developers focused their attention on the anti-cheat system. This feature allows players to check whether they have met cheaters in their last matches. The system detects not all types of cheating, but only certain things, like camera scale or items, drop before using some items, like Arcane Boots or Soul Ring.

In fact, it is not an anti-cheat as it is. It is not punishing the violators, but only notices some types of cheating. Soon after this feature was released, a wave of bans occurred, under which many cheaters were put out of the game. The release of this feature by GOSU.AI could affect this, but most likely is just a coincidence.

The analysis of the match by Divine 5+ player, which costs 400 gold or 16 dollars.

It starts with the pick. An expert is evaluating how good was the choice of the hero in this exact match-up and suggesting alternative picks. After that he analyzes the game from the very beginning, pointing out mistakes and correct decisions. Somewhere he suggests to pay more attention to the last hitting and balancing the creeps on the lane, somewhere he advises to communicate more with the teammates for better cooperation, comments on the choice of the items and pointing out the alternatives. Also, analytic is stating the mistakes in fights like wrong ultimate usage or contact with skills like Power Cogs. In the last section, the expert notes “global mistakes” and giving a few tips on how to improve in macro. In my match, analytic paid most of the attention to four things

Lasthitting on the early stages
Power Treads switching
Poor usage of the ultimate on the early stages
Poor team communication

Analytic recommended me to watch the replays of professional players, individually train the last hit and communicate with the team more.

Since reviews are not done automatically, it is difficult to evaluate the quality as a whole. There may be two completely different matches, so as two completely different analyses of them. The example that was on the site was much more detailed than the analysis that I received, while my match lasted longer by 20 minutes, and my hero participated in more team fights. Yes, perhaps I made fewer mistakes and made more correct decisions — but in this case, the review does not reflect this. In 18 of 19 cases, the analyst told me about my mistakes. There is a lack of something like a game plan, based on the pick. Our heroes are better in the early stages, then we should go push? Or vice versa, it was important not to make mistakes on the lanes, as enemies would put bigger pressure in first 20 minutes, so this exact death ruined everything? What are the goals of my team or my hero’s personal goals, what are the most significant timings? Some of those elements are present in the basic analysis, but it would be cool to see them in the separate section

An analysis, used as an example on the website

The review also lacks a conclusion. This would be useful when discussing the reasons of the defeat, like: “On 20 minute your team had 10K gold advantage, and you did not take advantage of it, everyone was just farming, even though your pick was worse in the late game. That’s why, at some moment, your opponents became stronger”. Suck type of analysis was not present in both, example on the website and my personal analysis.

In my personal opinion, this analysis doesn’t worth the money paid. An expert is pointing out some interesting moments, which may teach you something, however, there are too many details, which make the overall picture blur out. Of course, maybe I received an unfortunate option, but if we consider it as an educational instrument and also consider the price, I would better spend a bit more and took personal coaching. Even one hour with a good coach may be much more useful.

The second feature, which is available only for gold, is not educational. GOSU.AI collects the user’s highlights of the match. It cost 50 gold, and the video turns out a bit strange. Few moments, team fights are cut off, there are commentaries on the screen where Phase Boots are called “Exciting”. Moreover, this music, which makes it look more like a parody on real highlight compilation.

As before, GOSU.AI is not a good teaching service. The information it provides appears limited. To improve your skill you have to make conclusions yourself. For sure, working on yourself is normal for education. However, in this case, it is simply better to watch replays, as there is more information and it remains in context, and use GOSU.AI as a searching resource for your highlights

MoreMMR

MoreMMR website launched in August 2017. Developers engaged professionals in their project but did not announce names, except Roman Resolut1on.

Previously, the platform appeared as an educational resource with five chains of quests, similar to Compendium quests. A player watched a video, which explained some game element, then he had to perform a task, to go on the certain lane on a certain hero, last hit some amount of creeps and perform a kill. After which a player would get access to next guide.

Now, portal got rid of the rigid structure and was supplemented by a new function. Now, it is a collection of video guides like Gameleap and PVGNA.

The features added:

There are two types of videos on the website, which are guides on heroes and general content. Most of the videos on certain heroes are based on public matches of top players How to lane, what to do in the midgame and which items to get. Generally, it is interesting and useful to watch, especially if you do not play on the hero you are watching a guide about. Developers are regularly showing the situations, which a player perform under pressure on the lane, so there is something to learn from.

The match analysis on this platform is not too descriptive, usually, it simply explains what and why the player is doing. This may be useful for players, who are not that familiar with a hero. More experienced players can watch the replays, you do not necessarily need a website for it.

Videos in “General” section are explaining the essentials of a certain position, hero pick, proper warding and other themes, which relate to multiple characters. The system of quests has also been changed. Quests are now not necessary, and if before they were attached to certain heroes, now there are certain quests for every hero. After a player completes the quest, he gets gold, which could be spent on a subscription, which gives you access to more guides.

Each task requires certain conditions to be fulfilled, like deal X amount of physical or magical damage, place X amount of wards, finish the match with certain KDA, participate in X amount of team fights, kill X amount of enemies in a row and purchase certain items before X minute. If you did not fulfill at least one of the conditions of the quest, you will not complete the quest. Some quests appear as really complicated.

There are 3 types of subscription on the website: Free, Silver Pro and Gold Pro. Depending on the type of it, the user gets different parts of the content. The first one is free, Silver and Gold worth 9 and 18 gold a day. It equals to 9 and 18 cents, respectively. After a player completes a quest, he gets 1, 3 and 6 gold. For all the quests for one hero a player can get from 40 to 60 gold, depends on how many quests particular hero has.

During the first revision of MoreMMR, we wondered how often developers are willing to update the content to keep up with new patches. Now it became clear, the content is updating, but it goes relatively slowly. Seems like the basic videos with themes like “The role of support in game” are updated pretty regularly, more specific ones are lagging behind. However, the average of around 30 videos a week gives hope for more and more relevance. All the videos are marked with a patch number it is relative to.

Also, the portal presented the match analysis. It simply tells a player, which elements of his performance have mostly lowered his chances to win. Not enough farming or EXP, bad last hit, not enough damage to the buildings, or advising on picking most meta-hero. Those are the parameters on which the win depends, according to MoreMMR developers. It is hard to judge on the objectivity of those parameters, as there are different games, some heroes’ playstyle simply does not require hitting towers, and some unpopular heroes may be a good counter in particular match.

Generally, the match analysis function cannot be called perfect right now. Let’s say I played Lich as an easy lane support. So why MoreMMR is claiming that I played carry, telling me how low was my last hit stat and how bad I was at hitting buildings. Yes, I did a mistake by taking +120 DMG talent on LvL 15, but that’s all I had similar to carry position. Yes, a hero has different talents for this position, but it is not a reason. Moreover, even a hero like Oracle, who definitely do not have such talents, is determined as by the platform as a carry.

However, developers regularly report on hotfixes and updates of the match analysis, and about new mistakes and bugs found.

The main problem with the platform in the past was the cost. It was possible to find cheaper or even free content of similar quality. Now guides are much cheaper, as the subscription model was implemented, and the user is provided with some new free features, like free stats on heroes, match analysis, and Weekly Cups. The platform now is much more flexible, and a highly skilled user can choose for himself, which content he is interested in and do not waste time on going through things he already knows. There is more to the new free functions.

Recently, MoreMMR added new entertaining feature “Weekly Cup” — a free competition for everyone with prizes. In fact, these are the same quests, but with harder missions. Five quests each week, where 5 users who completed it in the fastest time, receive Arcanas. The timer is individual for every player. For now, the feature is being tested, some games are being checked for a relatively long time, but in two first cups, they had more than 1000 players in each cup.

Feedless

Feedless is a companion program that provides players with some tips in English. It informs a player about missing enemy heroes, how soon runes will appear, reminds a player when his ward will drop and when to stack a jungle camp. Most of the functions are available in the free version, and improved version of Feedless has such features as reminding a player that he has to carry a TP scroll in the backpack and that wards remain there too long. It also helps to choose items, informing on the possibility of being ganked and that enemies will make a kill attempt on a player soon.

Feedless has an interesting concept — during the game, the program takes screenshots, sends them to the server, analyzes them and gives advice.

The program works in overlay and turns on the picking stage. The window displays heroes, which work great with heroes your team already picked and counters enemy heroes. A player can set up the desired role and lane, to get more descriptive advice. In other sections a player can check up how, most likely, enemy heroes will line up, which heroes your teammates and opponents played recently and how, in theory, your and enemy’s pick is strong on various game stages. The interface is incomprehensible, at first it is hard to understand it fast in the limited time of the pick. Program lack some kind of a preview, pre-game explanation or at least a guide on the website.

At this point, I found Feedless useless. You would rarely find logic in best heroes suggested, and that advice were more distractive, rather than helpful. The principle on which the program decides counterpicks and hero synergies is not obvious. Developers are stating that they applied “principles of machine learning, based on the experience of 500K matches”. The situation with such type of analysis shows why particular situations in Dota are more important, that statistics.

After the picking stage ends, the window disables and Feedless starts to talk with the player. The program, as previously said, is advising on possible ganks, wards, informs on runes and stacks, reminds of TP scroll and etc. In fact, that all could be tracked by the player himself if he would check the time and the minimap. Feedless is not providing with something exceptionally special, it is not a cheat.

When the program advised my Anti Mage to go gank another lane on lvl 3, it became obvious that it’s advice are far from reality. Maybe, I got such advice as enemies were in certain position on the lane, but lvl 3 Anti Mage is not the hero for ganking. However, the program has lots of settings, every type of advice could be disabled, leaving only what you personally need. It says when Roshan may appear when Aegis will expire, is there a ward or book available in the store. Useful, but those are things you can personally keep track of, the most important is to click on a timer in time and do not flood in the chat. Item suggestion may only help those players, who are not that familiar with the hero they picked, but in-game guides on items are also doing this job pretty well.

High skilled players will not find Feedless helpful. They are experienced enough to follow up with the minimap, the timings of Rosh and Aegis, detect targets for ganks and push, and, of course, remember stacks and rune timings.

PVGNA

Pvgna is a resource with video guides in English. Content is ranged in themes: How to upgrade in rank, how to play a certain role, how to win in patch 7.06. and etc. There is also a section on basic mechanics and concepts. Those branches consist of guides with more narrow topics, firstly there are basics, and then there are more descriptive guides. Every guide is divided into shorter videos, averagely 15 minutes each. Some videos do not belong to any theme.

The website lacks a bit more concrete structure. When registering, a player is asked to fill in his MMR, favorite heroes and those heroes which you would like to learn. But if there is no clear understanding of what you would like to work on, it is hard to navigate. It may be hard for a player to understand on how to properly approach such amount of information, which guide to choose, what is worth of primary look and how to process so much information and structure it in your head. However, if you have a determined goal, Pvgna definitely would be very useful.

At the beginning of summer, Pvgna has launched Oracle. It is a free service on personal match analysis, or, as the developers call it “The Dota 2 coach with artificial intelligence”. The program evaluates a player based on 6 variables:

Mobility -Survivability
Farming -Objects
Teamfights -Vision

The system compares those with the average level of player’s MMR. Every variable has subvariables, the graphs look cool and descriptions, how it is important and why, and also how this variable is handled by higher MMR players. For example, the parameter of survivability is combined of how often a player uses Magic Wand, the number of deaths, how much time is spent in the tavern, how often he survives team fights and what is player’s percent of gold lost, because of deaths. A player can specify his role, then the characters in those variables would change, as the program measures carry and support positions based on different categories.

The exact match analysis from Oracle is less descriptive. It measures how good both teams performed in those 6 parameters, and how every player performed on his own role. The analysis lacks a bit of explanation on why this way and what could be improved.

Irrelevant to your current role, Oracle can give you pretty objective, and part of parameters is evaluated based on team results, not personal performance. Also, when choosing the roles, variables are changing. This info may give you an understanding of your weaknesses and lead you to conclusions on areas for improvements. The most important thing is to not to get too involved in statistics and apply common sense, as sometimes it is better to risk and win, even though it may affect your stats.

GameLeap

GameLeap is another English-speaking portal with video guides. The level of authors here is higher. They are either players with 6–7K+ MMR, or professionals. A resource has more solid structure. Content is divided into two categories. First one is general guides and manuals for heroes. Here you will see information on mechanics, roles, psychology in Dota, look at how high lvl players analyze replays. In another category of pro players there are 101 guides, averagely they are 1.5–2 hours long. The longest one is Arc Warden guide which is 6 hours long. Unlike Pvgna, guides here do not duplicate, usually it is one hero per one guide, however, some heroes have different guides for different in-game roles.

Every guide is divided in thematic parts. Themes are chosen by authors and differ from each other, so the quality and usefulness range from guide to guide. One author may pay more attention to the beginning stage of the game, another one would focus more on proper decision making, the third author would take a half an hour explaining when to pick this hero. Some guides are even not finished and consist of only intro and a couple of points. An interphase in a bit inconvenient. For example, there is no option to check all guides as a table. “Cards” of heroes look very cool, but not so informative

From the first look at the content it is noticeable that authors have different approach to the same tasks. However, some content creators go a bit too far with theory, without much illustrations. Within the authors there is one who really outstands, it is Brian “BananaSlamJamma” Canavan. Probably, his experience of individual coaching helped to understand, which questions and problems may be relevant for players. That’s why he is really good in clearly in explaining his point of view. He is exceptionally good in content for carry heroes, in both, general guides so as with certain heroes.

Gameleap and Pvgna have one thing in common, it is the fact that part of their content may be found outside of the platform. It is available on streams of professionals, replays, search for an exact video on YouTube. However, you are paying in order not to spend time searching for that information and get the guidance you need in the fast and convenient way. You have to always clearly understand, what do you want. Only then those websites would be maximum efficiency. Especially it is noticeable on Pvgna. There is more content there, and it is partially more in-deep, but the resource lacks structure, and that’s why a user may easily drown in such big cluster of information.