54:. In AI, two-player zero-sum games (such as heads-up hold'em) are usually won by approximating a Nash equilibrium strategy; however, this approach does not work for games with three or more players. Pluribus instead uses an approach which lacks strong theoretical guarantees, but nevertheless appears to work well empirically at defeating human players. Across the competitions, Pluribus won an average of over 30 milli big blinds per game. Pluribus' self-learned play style avoids "limping" (calling the big blind), and engages in "donk betting" (ending a round with a call and starting the next round by betting) more often than human experts do.
49:
According to the
Pluribus creators, "Developing a superhuman AI for multiplayer poker was the widely recognized main remaining milestone" in computer poker prior to Pluribus. Pluribus relies on offline self-play to build a base strategy, but then continues to learn in real-time during its online
61:
stated "Pluribus is a very hard opponent to play against. It's really hard to pin him down on any kind of hand." Jimmy Chou stated "Whenever playing the bot, I feel like I pick up something new to incorporate into my game." In
71:
Playing No-Limit Hold'em against five professional poker players, Pluribus won an average of $ 5 per hand with winnings of $ 1,000 per hour, which
Facebook described as a "decisive margin of victory."
50:
play. The base strategy was computed in eight days, and at market rates would cost about $ 144 to produce, much smaller than contemporary superhuman game-playing milestones such as
74:
Following the victory, the developers declined to release the source code, out of fear it would be misused to surreptitiously cheat against human poker players in online matches.
324:
305:
170:
92:
46:
and is "the first bot to beat humans in a complex multiplayer competition". The developers of the bot published their results in 2019.
196:
348:
57:
Among expert poker players, Jason Les stated he felt "very hopeless. You don't feel like there’s anything you can do to win."
35:
279:
64:
27:
68:, science editor Daniela Hernandez characterized Pluribus as "advanced at a key human skill — deception".
23:
234:
133:
325:"Facebook's new poker-playing AI could wreck the online poker industry—so it's not being released"
260:
252:
151:
43:
242:
141:
238:
137:
58:
342:
264:
306:"Facebook AI Pluribus defeats top poker professionals in 6-player Texas Hold 'em"
280:"Facebook, Carnegie Mellon build first AI that beats pros in 6-player poker"
247:
222:
146:
121:
51:
256:
155:
31:
93:"This Poker-Playing A.I. Knows When to Hold 'Em and When to Fold 'Em"
39:
197:"Computers Can Now Bluff Like a Poker Champ. Better, Actually"
171:"Facebook and CMU's 'superhuman' poker AI beats human pros"
120:Brown, Noam; Sandholm, Tuomas (11 July 2019).
8:
246:
145:
115:
113:
83:
221:Brown, Noam; Sandholm, Tuomas (2019).
223:"Superhuman AI for multiplayer poker"
122:"Superhuman AI for multiplayer poker"
7:
304:Ouellette, Jennifer (11 July 2019).
195:Hernandez, Daniela (11 July 2019).
14:
169:Vincent, James (11 July 2019).
91:Solly, Meilan (15 July 2019).
1:
323:Knight, Will (11 July 2019).
278:Brown, Noam (11 July 2019).
365:
36:Carnegie Mellon University
248:10.1126/science.aay2400
147:10.1126/science.aay2400
65:The Wall Street Journal
28:artificial intelligence
349:Computer poker players
329:MIT Technology Review
38:. Pluribus plays the
24:computer poker player
16:Computer poker player
239:2019Sci...365..885B
201:Wall Street Journal
138:2019Sci...365..885B
42:variation no-limit
233:(6456): 885–890.
132:(6456): 885–890.
356:
333:
332:
320:
314:
313:
301:
295:
294:
292:
290:
275:
269:
268:
250:
218:
212:
211:
209:
207:
192:
186:
185:
183:
181:
166:
160:
159:
149:
117:
108:
107:
105:
103:
88:
364:
363:
359:
358:
357:
355:
354:
353:
339:
338:
337:
336:
322:
321:
317:
303:
302:
298:
288:
286:
284:ai.facebook.com
277:
276:
272:
220:
219:
215:
205:
203:
194:
193:
189:
179:
177:
168:
167:
163:
119:
118:
111:
101:
99:
90:
89:
85:
80:
17:
12:
11:
5:
362:
360:
352:
351:
341:
340:
335:
334:
315:
296:
270:
213:
187:
161:
109:
82:
81:
79:
76:
59:Chris Ferguson
44:Texas hold 'em
15:
13:
10:
9:
6:
4:
3:
2:
361:
350:
347:
346:
344:
330:
326:
319:
316:
311:
307:
300:
297:
285:
281:
274:
271:
266:
262:
258:
254:
249:
244:
240:
236:
232:
228:
224:
217:
214:
202:
198:
191:
188:
176:
172:
165:
162:
157:
153:
148:
143:
139:
135:
131:
127:
123:
116:
114:
110:
98:
94:
87:
84:
77:
75:
72:
69:
67:
66:
60:
55:
53:
47:
45:
41:
37:
33:
29:
25:
21:
328:
318:
310:Ars Technica
309:
299:
287:. Retrieved
283:
273:
230:
226:
216:
204:. Retrieved
200:
190:
178:. Retrieved
174:
164:
129:
125:
100:. Retrieved
96:
86:
73:
70:
63:
56:
48:
19:
18:
289:23 February
102:23 February
97:Smithsonian
34:AI Lab and
206:4 February
180:4 February
78:References
32:Facebook's
265:195892791
175:The Verge
52:AlphaZero
30:built by
343:Category
257:31296650
156:31296650
20:Pluribus
235:Bibcode
227:Science
134:Bibcode
126:Science
263:
255:
154:
26:using
261:S2CID
40:poker
22:is a
291:2023
253:PMID
208:2021
182:2021
152:PMID
104:2023
243:doi
231:365
142:doi
130:365
345::
327:.
308:.
282:.
259:.
251:.
241:.
229:.
225:.
199:.
173:.
150:.
140:.
128:.
124:.
112:^
95:.
331:.
312:.
293:.
267:.
245::
237::
210:.
184:.
158:.
144::
136::
106:.
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.