ANAC Competition¶
The Automated Negotiating Agents Competition (ANAC) is an international competition that has been running since 2010 to bring together researchers from the negotiation community. The competition challenges participants to design autonomous negotiation agents that can effectively negotiate with other agents in various scenarios.
NegMAS provides comprehensive support for ANAC competitions through the
ANAC_INFO (also available as GENIUS_INFO) dictionary, which contains
metadata about all ANAC competition years from 2010 to present. The ANAC
community is committed to advancing he field through open-sourcing ALL
strategies.
Note
The competition had one League and ran on the Java-based Genius platform. Starting in 2020, the main league began using the GeniusWeb platform, and from 2024, it was named the Automated Negotiation League (ANL) and runs natively on NegMAS.
Accessing Competition Data¶
You can access information about any competition year using the ANAC_INFO or
ANAC_INFO dictionaries:
from negmas.genius.ginfo import ANAC_INFO, ANAC_INFO, get_anac_agents
# ANAC_INFO is an alias for ANAC_INFO
assert ANAC_INFO is ANAC_INFO
# Get information about a specific year
info_2024 = ANAC_INFO[2024]
print(info_2024["winners"]) # List of winners
print(info_2024["finalists"]) # List of finalists
print(info_2024["participants"]) # All participants
# Use helper function to get agents with specific criteria
winners_2024 = get_anac_agents(year=2024, winners_only=True)
bilateral_agents = get_anac_agents(bilateral=True)
Agent Availability¶
Agents from different competition years are available through different libraries:
- Java/Genius Agents (2010-2019)
These agents were written in Java for the Genius platform. To use them, you need:
GeniusBridge: A Java bridge that must be running to communicate with Java agents. See the Integrating with Genius tutorial for setup instructions.
negmas-genius-agents (optional): A library that re-implements many of these agents in pure Python using AI-based conversion. This allows running them without the Java bridge. Install with:
pip install negmas-genius-agents
- GeniusWeb Java Agents (2020-2021)
These agents were written in Java for the GeniusWeb platform. To use them:
Install negmas-geniusweb-bridge:
pip install negmas-geniusweb-bridgeThe bridge provides AI-translated Python implementations of the Java agents.
All agents are available as GW-prefixed wrapped classes (e.g.,
GWAlphaBIU)
- GeniusWeb Python Agents (2022-2023)
These agents were written in Python for the GeniusWeb platform. To use them:
Install negmas-geniusweb-bridge:
pip install negmas-geniusweb-bridgeAll agents are available as GW-prefixed wrapped classes (e.g.,
GWExploitAgent)
- ANL Agents (2024+)
These agents are written in pure Python for NegMAS. To use them:
Install anl-agents:
pip install anl-agents
Platform Compatibility¶
The following table shows which platforms were used for each competition year and how to access the agents in NegMAS:
Years |
Official Platform |
NegMAS Compatible Implementation |
Notes |
|---|---|---|---|
2010-2019 |
Genius (Java) |
GeniusBridge + |
Requires Java runtime and GeniusBridge running. See Integrating with Genius. |
2010-2019 |
Genius (Java) |
AI-translated Python implementations. Install: |
|
2020-2021 |
GeniusWeb (Java) |
AI-translated from Java. 6 agents from 2021 are available. Install: |
|
2022-2023 |
GeniusWeb (Python) |
Native Python agents wrapped for NegMAS. Install: |
|
2024+ |
NegMAS (Python) |
Native NegMAS agents. Install: |
Competition Years¶
Below is a detailed summary of each competition year, including the settings, winners, and agent availability.
2010¶
Settings: Bilateral, Linear utilities, Discounting
Winners:
AgentK (
agents.anac.y2010.AgentK.Agent_K)Yushu (
agents.anac.y2010.Yushu.Yushu)Nozomi (
agents.anac.y2010.Nozomi.Nozomi)IAMhaggler (
agents.anac.y2010.Southampton.IAMhaggler)
Participants (7 agents): AgentFSEGA, AgentK, AgentSmith, Nozomi, IAMcrazyHaggler, IAMhaggler, Yushu
Agent Access: Java/Genius (requires GeniusBridge)
2011¶
Settings: Bilateral, Linear utilities, Discounting
Winners:
HardHeaded (
agents.anac.y2011.HardHeaded.KLH)Gahboninho (
agents.anac.y2011.Gahboninho.Gahboninho)IAMhaggler2011 (
agents.anac.y2011.IAMhaggler2011.IAMhaggler2011)
Finalists (8 agents): HardHeaded, Gahboninho, IAMhaggler2011, BramAgent, AgentK2, TheNegotiator, Nice Tit-for-Tat, ValueModelAgent
Agent Access: Java/Genius (requires GeniusBridge)
2012¶
Settings: Bilateral, Linear utilities, Reservation value, Discounting
Winners:
CUHKAgent (
agents.anac.y2012.CUHKAgent.CUHKAgent)AgentLG (
agents.anac.y2012.AgentLG.AgentLG)OMACagent (
agents.anac.y2012.OMACagent.OMACagent)
Finalists (8 agents): CUHKAgent, AgentLG, OMACagent, AgentMR, TheNegotiatorReloaded, BRAMAgent2, MetaAgent, IAMhaggler2012
Agent Access: Java/Genius (requires GeniusBridge)
2013¶
Settings: Bilateral, Linear utilities, Learning, Reservation value, Discounting
Winners:
TheFawkes (
agents.anac.y2013.TheFawkes.TheFawkes)MetaAgent2013 (
agents.anac.y2013.MetaAgent.MetaAgent2013)TMFAgent (
agents.anac.y2013.TMFAgent.TMFAgent)
Finalists (7 agents): TheFawkes, MetaAgent2013, TMFAgent, AgentKF, InoxAgent, Slavaagent, GAgent
Agent Access: Java/Genius (requires GeniusBridge)
2014¶
Settings: Bilateral, Non-linear utilities, Reservation value
Winners:
AgentM (
agents.anac.y2014.AgentM.AgentM)DoNA (
agents.anac.y2014.DoNA.DoNA)Gangster (
agents.anac.y2014.Gangster.Gangster)
Finalists (9 agents): AgentM, DoNA, Gangster, AgentYK, BraveCat, E2Agent, Atlas3, AgentTRP, WhaleAgent
Participants (18 agents): Full list in ANAC_INFO[2014]["participants"]
Agent Access: Java/Genius (requires GeniusBridge)
2015¶
Settings: Multilateral, Linear utilities, Reservation value
Winners:
Atlas3 (
agents.anac.y2015.Atlas3.Atlas3)ParsAgent (
agents.anac.y2015.ParsAgent.ParsAgent)RandomDance (
agents.anac.y2015.RandomDance.RandomDance)
Finalists (8 agents): Atlas3, ParsAgent, RandomDance, AgentX, Kawaii, Mercury, PhoenixParty, PokerFace
Participants (24 agents): Full list in ANAC_INFO[2015]["participants"]
Agent Access: Java/Genius (requires GeniusBridge)
2016¶
Settings: Multilateral, Linear utilities, Reservation value
Winners:
Caduceus (
agents.anac.y2016.caduceus.Caduceus)YXAgent (
agents.anac.y2016.yxagent.YXAgent)
Finalists (10 agents): Caduceus, YXAgent, ParsCat, Farma, Atlas32016, MyAgent, Ngent, GrandmaAgent, AgentHP2, Terra
Participants (16 agents): Full list in ANAC_INFO[2016]["participants"]
Agent Access: Java/Genius (requires GeniusBridge)
2017¶
Settings: Multilateral, Linear utilities, Learning, Reservation value
Winners:
PonPokoAgent (
agents.anac.y2017.ponpokoagent.PonPokoAgent)CaduceusDC16 (
agents.anac.y2017.caduceus.CaduceusDC16)Rubick (
agents.anac.y2017.rubick.Rubick)
Finalists (10 agents): PonPokoAgent, CaduceusDC16, Rubick, AgentF, AgentKN, GeneKing, Mamenchis, ParsCat2, TucAgent, SimpleAgent2017
Participants (18 agents): Full list in ANAC_INFO[2017]["participants"]
Agent Access: Java/Genius (requires GeniusBridge)
2018¶
Settings: Multilateral, Linear utilities, Learning, Reservation value, Discounting
Winners:
MengWan (
agents.anac.y2018.meng_wan.Agent36)BetaOne (
agents.anac.y2018.beta_one.Group2)AgentHerb (
agents.anac.y2018.agentherb.AgentHerb)
Finalists (13 agents): MengWan, BetaOne, AgentHerb, ConDAgent, ExpRubick, FullAgent, IQSun2018, Libra, PonPokoRampage, Seto, Shiboy, SMAC_Agent, Yeela
Participants (20 agents): Full list in ANAC_INFO[2018]["participants"]
Agent Access: Java/Genius (requires GeniusBridge)
2019¶
Settings: Bilateral, Linear utilities, Reservation value, Uncertainty
Platform: Genius (Java)
Winners:
AgentGG (
agents.anac.y2019.agentgg.AgentGG)KakeSoba (
agents.anac.y2019.kakesoba.KakeSoba)SAGA (
agents.anac.y2019.saga.SAGA)
Finalists (6 agents): AgentGG, KakeSoba, SAGA, DandikAgent, GaravelAgent, MINF
Participants (18 agents): Full list in ANAC_INFO[2019]["participants"]
Agent Access: Java/Genius (requires GeniusBridge)
2020¶
Settings: Linear utilities, Reservation value, Uncertainty, Elicitation
Platform: GeniusWeb (Java)
Finalists (5 agents, alphabetically - no explicit ranking published):
AgentKT (Brown University, USA)
AhBuNeAgent (Ozyegin University, Turkey)
Angel (University of Tulsa, USA)
HammingAgent (TUAT, Japan)
ShineAgent (Bar-Ilan University, Israel)
Participants (13 agents): AgentKT, AgentP1DAMO, AgentXX, AhBuNeAgent, Anaconda, Angel, AzarAgent, BlingBling, DUOAgent, ForArisa, HammingAgent, NiceAgent, ShineAgent
Agent Access: negmas-geniusweb-bridge
2021¶
Settings: Linear utilities, Learning, Reservation value, Uncertainty, Elicitation
Platform: GeniusWeb (Java)
Winners:
AlphaBIU (Bar-Ilan University, Israel)
MatrixAlienAgent (University of Tulsa, USA)
TripleAgent (Utrecht University, Netherlands)
Agent Access: negmas-geniusweb-bridge
Participants (6 agents): AgentFO2021, AlphaBIU, GamblerAgent, MatrixAlienAgent, TheDiceHaggler2021, TripleAgent
Agent Access: negmas-geniusweb-bridge
Note
Six ANAC 2021 agents have been AI-translated from Java and are available in the bridge package. Additional agents may be added in future releases.
2022¶
Settings: Linear utilities, Learning, Reservation value, Elicitation
Platform: GeniusWeb (Python)
Winners (Individual Utility):
DreamTeam109Agent (College of Management Academic Studies, Israel)
ChargingBoul (University of Tulsa, USA)
Winners (Social Welfare):
DreamTeam109Agent (College of Management Academic Studies, Israel)
Agent007 (Bar-Ilan University, Israel)
Finalists (3 agents): Agent007, ChargingBoul, DreamTeam109Agent
Participants (19 agents): Agent007, Agent4410, AgentFish, AgentFO2, BIU_agent, ChargingBoul, CompromisingAgent, DreamTeam109Agent, GEAAgent, LearningAgent, LuckyAgent2022, MiCROAgent, Pinar_Agent, ProcrastinAgent, RGAgent, SmartAgent, SuperAgent, ThirdAgent, Tjaronchery10Agent
Agent Access: negmas-geniusweb-bridge
2023¶
Settings: Linear utilities, Learning, Reservation value, Elicitation
Platform: GeniusWeb (Python)
Winners (Individual Utility):
ExploitAgent - Bram Renting (Leiden University, Delft University of Technology)
MiCRO2023 - Dave de Jonge (IIIA-CSIC)
Winners (Social Welfare):
AntHeartAgent - Kaiyou Lei et al. (Ant Group)
SmartAgent - Jianing Zhao et al. (Southwest University)
Finalists (4 agents): ExploitAgent, MiCRO2023, AntHeartAgent, SmartAgent
Participants (15 agents): AgentFO3, AmbitiousAgent, AntAllianceAgent, AntHeartAgent, ColmanAnacondotAgent2, ExploitAgent, GotAgent, HybridAgent2023, KBTimeDiffAgent, MiCRO2023, MSCAgent, PopularAgent, SmartAgent, SpaghettiAgent, TripleEAgent
Agent Access: negmas-geniusweb-bridge
2024 (ANL)¶
Settings: Bilateral, Linear utilities, Reservation value, Known opponent utility function
Platform: NegMAS (Native Python)
Winners (Individual Advantage):
Shochan - Takayama, TUAT, Japan (
anl_agents.anl2024.takafam.Shochan)UOAgent - Hirotada Matsumoto, TUAT, Japan (
anl_agents.anl2024.team_moto.UOAgent)AgentRenting2024 - Mick Elshout, Utrecht University, Netherlands (
anl_agents.anl2024.team_renting.AgentRenting2024)
Winner (Nash Optimality):
Shochan - Takayama, TUAT, Japan
Finalists (10 agents): Shochan, UOAgent, AgentRenting2024, AntiAgent, HardChaosNegotiator, KosAgent, Nayesian2, CARCAgent, BidBot, AgentNyan
Participants (19 agents): Full list in ANAC_INFO[2024]["participants"]
Agent Access: anl-agents (pip install anl-agents)
More Information: ANL 2024 Results
2025 (ANL)¶
Settings: Bilateral, Linear utilities, Reservation value, Multi-deal
Platform: NegMAS (Native Python)
Winners:
1. RUFL - Garrett Seo, Rutgers University, USA (anl_agents.anl2025.team_271.RUFL) - 250€
1. SAC Agent - Hossein Savari, University of Tehran, Iran (anl_agents.anl2025.university_of_tehran.SacAgent) - 250€ (tied for 1st)
3. UFunAtAgent - Fukutoku Yuma, TUAT, Japan (anl_agents.anl2025.team_305.UfunATAgent) - 100€
Finalists (12 agents, by qualification score):
SacAgent (University of Tehran)
ProbaBot (CWI, Netherlands)
RUFL (Rutgers University)
KDY (TUAT)
JeemNegotiator (Bar-Ilan University)
Astrat3m (Chongqing Jiaotong University)
A4E (TUAT)
OzUAgent (Ozyegin University)
SmartNegotiator (Bar-Ilan University)
CARC2025 (HIT Shenzhen)
UfunATAgent (TUAT)
Wagent (Chongqing Jiaotong University)
Participants (17 agents): Full list in ANAC_INFO[2025]["participants"]
Agent Access: anl-agents (pip install anl-agents)
More Information: ANL 2025 Results
Competition Settings Reference¶
The following table summarizes the settings for each competition year:
Year |
Type |
Linear |
Learning |
Reservation |
Discount |
Uncertainty |
Elicitation |
Multi-deal |
Known Opp. Ufun |
Known Opp. RV |
Platform |
|---|---|---|---|---|---|---|---|---|---|---|---|
2010 |
Bilateral |
Yes |
No |
No |
Yes |
No |
No |
No |
No |
No |
Genius |
2011 |
Bilateral |
Yes |
No |
No |
Yes |
No |
No |
No |
No |
No |
Genius |
2012 |
Bilateral |
Yes |
No |
Yes |
Yes |
No |
No |
No |
No |
No |
Genius |
2013 |
Bilateral |
Yes |
Yes |
Yes |
Yes |
No |
No |
No |
No |
No |
Genius |
2014 |
Bilateral |
No |
No |
Yes |
No |
No |
No |
No |
No |
No |
Genius |
2015 |
Multilateral |
Yes |
No |
Yes |
No |
No |
No |
No |
No |
No |
Genius |
2016 |
Multilateral |
Yes |
No |
Yes |
No |
No |
No |
No |
No |
No |
Genius |
2017 |
Multilateral |
Yes |
Yes |
Yes |
No |
No |
No |
No |
No |
No |
Genius |
2018 |
Multilateral |
Yes |
Yes |
Yes |
Yes |
No |
No |
No |
No |
No |
Genius |
2019 |
Bilateral |
Yes |
No |
Yes |
No |
Yes |
No |
No |
No |
No |
Genius |
2020 |
Mixed |
Yes |
No |
Yes |
No |
Yes |
Yes |
No |
No |
No |
GeniusWeb (Java) |
2021 |
Mixed |
Yes |
Yes |
Yes |
No |
Yes |
Yes |
No |
No |
No |
GeniusWeb (Java) |
2022 |
Mixed |
Yes |
Yes |
Yes |
No |
No |
Yes |
No |
No |
No |
GeniusWeb (Python) |
2023 |
Mixed |
Yes |
Yes |
Yes |
No |
No |
Yes |
No |
No |
No |
GeniusWeb (Python) |
2024 |
Bilateral |
Yes |
No |
Yes |
No |
No |
No |
No |
Yes |
No |
NegMAS |
2025 |
Bilateral |
Yes |
No |
Yes |
No |
No |
No |
Yes |
No |
No |
NegMAS |
Legend:
Multi-deal: Whether agents negotiate multiple deals simultaneously (introduced in ANL 2025)
Known Opp. Ufun: Whether agents have access to their opponent’s utility function (ANL 2024 only)
Known Opp. RV: Whether agents know their opponent’s reserved value (not used in any year so far)
Acknowledgements¶
Compiling the complete list of agents for ANAC since 2010 would have been impossible without the support of many people including:
Catholijn Jonker (TU Delft)
Tim Baarslag (CWI / TU Eindhoven)
Reyhan Aydogan (Ozyegin University / TU Delft)
Mehmet Onur Keskin (Ozyegin University)
Bram Renting (Leiden University)
Wouter Pasman (TU Delft)
Tamara Florijn (TU Delft)
External Resources¶
anl-agents - ANL competition agents (2024+)
negmas-geniusweb-bridge - GeniusWeb agents wrapper (2020-2023)
negmas-genius-agents - Genius agents reimplemented in Python (2010-2019)