2015 Competition Results

Team DataLab Birds from the Czech Technical University in Prague did it again! They are the old and new AIBIRDS Champion. Congratulations to Tomas, Radim and Karel, very well done!

But this year it was not as easy for them to win as last year. The Grand Final between DataLab Birds and the runner up AngryHex from Italy and Austria was actually the most exciting AIBIRDS match ever. The leaderboard changed almost every minute, sometimes every few seconds. Even though the levels were very difficult, both AI agents did a great job in solving them, some with incredible shots. Datalab Birds solved all 8 levels, while AngryHex solved only 7 out of 8 levels, but with higher scores. Both agents were solving levels again and again in order to improve their overall score. Fortunately, both teams were not attending personally, otherwise there would have surely been one or two heart attacks among the team members. 

The Semi Final between DataLab Birds, AngryHex, IHSEV from France and Tori, the newcomer from Argentina, was also very close. The leaderboard changed frequently and in the closing minutes DataLab Birds were last until they solved another level and took the lead -- kicking Tori out of the Grand Final and into the honourable Third Place

The two Quarter Finals were relatively clear decisions, with some unfortunate casualties. In the 1st Quarter Final between IHSEV, PlanA+ from South Korea, s-Birds Returns from India and Tori, two teams s-Birds and PlanA+ got stuck in some levels. While PlanA+ could recover and moved on to different levels, s-Birds were not able to recover and kept looping. 

In the 2nd Quarter Final between DataLab Birds, AngryHex, and the two newcomers UFAngryBirdsC from Brazil and Adil from Germany, Adil was the unfortunate team who kept zooming in and out of the first level without actually making a shot. 

Here are the scores of the different rounds: 

Results after Qualification



Quarter Final 1








2. AngryHex



2. Tori


3. Datalab Birds



3. PlanA+


4. PlanA+



4. s-Birds Returns


5. S-Birds Returns



6. UFAngryBirdsC



7. Adil



Quarter Final 2


8. Tori*






1. DataLab Birds





2. AngryHex





3. UFAngryBirdsC





4. Adil



Semi Final


Grand Final



1. DataLab Birds



1. DataLab Birds


2. AngryHex



2. AngryHex


3. Tori






*Note: Team Tori had a problem with their AI agent in the qualification round. We ran their agent again after they fixed the problem. While they ended up having a higher score than Adil, we still ranked them as 8th for the purpose of determining the allocations to the quarter finals. 


Earlier in the day we had the first run of our new Competitive Track where AI agents are playing the same level with alternating shots. The agent who scores the winning shot gets all the points. Before playing a level, agents make a concealed bid of how much it is worth to them to make the first, respectively, second shot. The agent with the higher bid makes the first shot. If an agent who gets to make the shot it was bidding for wins the level, it has to pay the positive bid amount to the other agent. Agents play the same set of levels against all other agents. 

The Competitive Track Final was played by IHSEV and s-Birds Returns. We added two variants of the Naive Agent to the Final to make it more exciting. One of them always bid 0 and always shoots directly towards a random pig, while the other variant always bids half of what the pigs plus the remaining birds (assuming one bird per pig) are worth, again always shooting toward a random pig. 

While shooting at a random pig was very successful in the first AIBIRDS competition in 2012, it was not very successful in the first run of the Competitive Track. Both Naive Agent variants ended the Final with 0 points, while IHSEV, our 1st AIBIRDS Competitive Track Champion had 127610 points and the runner-up s-Birds Returns scored 61450 points. 


The 2015 Man vs Machine Challenge where we test if the best agents already beat humans ended like in previous years, with a human winner. Josef Bajada from Kings College London took the lead (341460) among the human players just a few minutes before the competition ended, leaving Nathan Sturtevant in second place (337890), Jingwei Xu in third (335820), Fabio Rafael Gallo in fourth (335580) and Santiago Ontanon in fifth place (334520). Many humans were able to solve all four levels, while the best agents only solved three levels. DataLab Birds were the best among the AI agents (225190), AngryHex second (180910) and IHSEV third (121360). Tori didn't solve any level. AngryHex almost solved the most difficult level and started with an amazing shot, but then couldn't see the last unprotected pig sitting in the grass. Their problem was that they used their own computer vision which couldn't distinguish the green pig from the green grass.