Datasets:

surry-hills-druid
/

zack-n-hasan

license: cc0-1.0
language:
  - en
pretty_name: 'Asmongold vs HasanAbi: Twitch Chat Logs (May 2026)'
size_categories:
  - 1M<n<10M
task_categories:
  - text-classification
  - other
tags:
  - twitch
  - chat-logs
  - social-media
  - political
  - gaming
  - community-analysis
configs:
  - config_name: hasanabi
    data_files:
      - split: train
        path: data/hasanabi/*.jsonl
  - config_name: zackrawrr
    data_files:
      - split: train
        path: data/zackrawrr/*.jsonl

zack-n-hasan

Seven days of live Twitch chat logs (7–14 May 2026) from two of the platform's largest and most politically opposite streamers: HasanAbi and Asmongold. Captured using Chatterino, the popular Twitch chat client.

Channels

Channel	Description
hasanabi	Political commentary and culture. One of Twitch's most-watched political streamers, with a progressive, democratic-socialist perspective.
asmongold	Gaming and culture. Asmongold's primary channel, focused on gaming content and broader commentary, with a culturally conservative-leaning audience.

Coverage

Period: 7 May 2026 – 14 May 2026 (8 log files per channel)
Capture tool: Chatterino 2.x
Timezone: GMT+1 (British Summer Time)
Total messages: ~1,043,367 chat messages across both channels

Interesting Stats

Stats generated by analyze_stats.py. Emote counts cover Unicode emoji and detected Twitch/7TV emotes. Bot accounts excluded from sender stats. All figures cover the 8-day capture window.

Metric	HasanAbi	Asmongold
Total messages	494,526	548,841
Unique chatters	29,571	30,774
Total emojis + emotes	396,756	212,032
Avg emojis/emotes per message	0.80	0.39
Avg message length	30 chars / 5 words	24 chars / 5 words
Avg word length	5 chars	4 chars
Vocabulary diversity (TTR)	0.1176	0.1080
Single-message chatters	30.5%	29.3%
Political affiliation	Progressive left	Culturally conservative-adjacent
Echo chamber	High	Moderate
Conscientiousness	Low individually, high collectively	Low individually and collectively
Openness	High within ideology, low outside it	Moderate on gaming, low on social change
Authoritarian leaning	Anti-authoritarian politics, authoritarian norms	Culturally authoritarian, low internal enforcement

HasanAbi

Total messages: 494,526 - avg ~61,816/day (partial days on 7th and 14th)
Unique chatters: 29,571
Total emojis + emotes: 396,756 - Unicode emoji 17,177, Twitch / 7TV emotes 379,579
- Twitch/7TV emotes are strongly associated with Twitch-native culture (Gen Z and younger millennials raised on the platform)
- Millennial text markers (lol, lmao, omg, ngl) appear in 4.1% of messages; Gen Z text markers (W, no cap, based, mid) in 0.7% - the low rate of both suggests expression routes through emotes rather than typed slang
Single-message chatters: 9,019 (30.5% of users sent exactly one message)
Repeat chatters: 20,552 (users who sent 2 or more messages)
Average messages per user: 16.72
Average message length: 30 characters / 5 words
Average word length: 5 characters
Average emojis/emotes per message: 0.80
Vocabulary diversity (type-token ratio, 100k token sample): 0.1176
Longest message: 500 characters
Most popular phrases (bigrams, excluding emote-only lines):
- i think (1,769), going to (1,442), the same (1,357), need to (1,326), all the (1,228), if you (1,179), as a (1,170), hasanabi i (1,532)
Most popular phrases (trigrams):
- a lot of (714), just subbed using (904), hasraid hasraid hasraid (856) - the last being a coordinated chat moment during viewer raids

Political affiliation: Progressive left / democratic socialist. Chat regularly references labour rights, Palestinian solidarity, anti-imperialism, universal healthcare, and anti-capitalist critique. Political framing is explicit and central to the channel's identity.

Language bias observations (regex-based detection; figures include counter-speech and quoting - treat as indicative, not definitive):

Intensity and anger: 25.5% of messages are all-caps, the highest of the two channels. This reflects a culture of performative intensity - much is emote-driven enthusiasm, but it may contribute to a higher-intensity register. Messages with multiple exclamation marks appear in 0.79% of messages.
Hate and dehumanising language: 0.29% of messages (1,447) contain kill/die/hate phrasing. Dehumanising language (vermin, parasite, subhuman) appears in 0.01% - rare, but present.
Sexism: Sexist terms appear in 0.24% of messages (1,193). Explicit misogyny markers (women are, female moment) appear in 0.02% - lower than Asmongold's chat.
Political outgroup hostility: 0.24% of messages (1,210) use political slurs (fascist, nazi, groomer, libtard) directed at perceived opponents. Given the channel's political identity, some of this is in-group shorthand for out-group figures rather than targeted harassment.
Antisemitic framing: Terms associated with antisemitic conspiracy framing (zionist, globalist, Jewish control) appear in 75 messages (0.02%). The channel's heavy focus on Israeli-Palestinian politics means some of these hits are legitimate political commentary rather than antisemitism; context-level review could help disambiguate.
Racial slurs: No significant detection across the 8-day window, consistent with active moderation.

Personality observations:

Echo chamber: Chat closely tracks the streamer's political positions. Approval signals (BASED, hasFIST) are constant; dissent is rare and quickly moderated. The ban and timeout rate in the logs is high.
In-group / out-group bias: Clear us-vs-them vocabulary - terms like bootlicker, liberal, and imperialist are used as out-group labels. In-group membership is confirmed through correct political framing.
Performative communication: The high emote-per-message ratio (0.80) suggests a significant portion of messages are reactive affirmations rather than substantive contributions.
Collective empathy: Chat expresses empathy toward cited victims of political events - though this tends to be ideological and collective rather than interpersonal.
Punitive moderation: The logs contain a large number of permanent bans, suggesting strict enforcement of community norms. Limited leniency appears to be extended to users who violate those norms.
Coordinated behaviour: Raid emote spam (hasraid, hasslam, hasbaited) shows strong coordinated participation during live events.
Conscientiousness: Low at the individual message level - impulsive reactions dominate - but the community shows collective conscientiousness around ideological consistency, with apparent effort to call out perceived contradictions in public figures.
Openness: High within its ideological framework. The chat engages readily with complex political and geopolitical topics. Openness drops sharply at the boundary of that framework - perspectives that challenge core positions are met with dismissal rather than engagement.
Agreeableness: High toward the streamer and toward politically aligned viewpoints. Low toward perceived opponents. Chatters who correct or challenge Hasan are tolerated if framed carefully; those who don't tend to disappear from the logs.
Authoritarian leaning: Apparently paradoxical. The political content is explicitly anti-authoritarian (anti-establishment, anti-imperialist), yet the community's internal behaviour is highly authoritarian - strict norm enforcement, heavy moderation, and low tolerance for dissent within the chat itself.
Relationship with the streamer: Chatters mention HasanAbi directly at a high rate - 45,601 messages (9.2% of all messages). The dominant mode is information-feeding: chatters bring news links, political context, and corrections to Hasan's attention, positioning themselves as co-participants in the political discussion rather than passive viewers. A secondary mode is approval-seeking - chatters share views they expect Hasan to endorse, fishing for validation. Outright sycophancy ("you've built an amazing community") is present but less common than collaborative engagement. Disdain directed at the streamer is rare and swiftly moderated.

Asmongold

Total messages: 548,841 - avg ~68,605/day (partial days on 7th and 14th)
Unique chatters: 30,774
Total emojis + emotes: 212,032 - Unicode emoji 9,729, Twitch / 7TV emotes 202,303
- Millennial text markers (lol, lmao, omg, ngl) appear in 7.5% of messages; Gen Z text markers (W, no cap, based, mid) in 1.4% - higher rates than HasanAbi's chat, indicating more expression through typed slang and less through emotes
- The millennial-to-gen-z marker ratio (~5:1) is similar across both channels, but Asmongold's overall higher text-slang rate suggests a slightly older or more cross-platform audience
Single-message chatters: 9,009 (29.3% of users sent exactly one message)
Repeat chatters: 21,765 (users who sent 2 or more messages)
Average messages per user: 17.83
Average message length: 24 characters / 5 words
Average word length: 4 characters
Average emojis/emotes per message: 0.39
Vocabulary diversity (type-token ratio, 100k token sample): 0.1080
Longest message: 500 characters
Most popular phrases (bigrams, excluding emote-only lines):
- i think (1,929), need to (1,768), the same (1,665), if you (1,549), going to (1,523), this guy (1,478), looks like (1,352), want to (1,421)
Most popular phrases (trigrams):
- a lot of (695), do you think (399), this guy is (396), looks like a (393), we need to (387), did you see (358), om om om (343)

Political affiliation: Culturally conservative-adjacent to libertarian-leaning. The chat is not primarily political, but where politics appears it skews toward scepticism of progressive positions on gaming, entertainment, and culture. Anti-DEI sentiment and resistance to political content in games are recurring themes.

Language bias observations (regex-based detection; figures include counter-speech and quoting - treat as indicative, not definitive):

Intensity and anger: 17.4% of messages are all-caps - lower than HasanAbi's chat despite a higher total message count. Multiple exclamation marks appear in 0.42% of messages. The lower caps rate suggests less performative intensity, though gaming reactions account for much of what is present.
Hate and dehumanising language: 0.38% of messages (2,075) contain kill/die/hate phrasing - a higher rate than HasanAbi's chat. Dehumanising terms appear in 0.01%. The gaming context accounts for some of this (in-game death references, competitive hostility), but the rate is comparatively higher than HasanAbi's.
Sexism: Sexist terms appear at a similar rate to HasanAbi's chat (0.23%, 1,254 messages). Explicit misogyny markers (women are, girl moment) appear in 0.04% - double the rate of HasanAbi's chat - consistent with a community where gender-essentialist framing appears more routinely.
Political outgroup hostility: 0.12% of messages (684) use political slurs, a lower rate than HasanAbi's chat, reflecting the channel's lower political engagement overall.
Antisemitic framing: Terms associated with antisemitic conspiracy framing appear in 286 messages (0.05%) - higher than HasanAbi's chat and less attributable to political commentary about a specific conflict.
Racial slurs: No significant detection across the 8-day window, consistent with active moderation.

Personality observations:

More varied discourse: Lower emote density (0.39 vs 0.80) and more interrogative bigrams (do you think, did you see) suggest a more varied pattern of back-and-forth than in the hasanabi chat.
Gaming-oriented identity: In-group membership is defined by gaming culture rather than political alignment. Outsiders who criticise gaming are treated with scepticism or mockery.
Reactive commentary: this guy is, looks like a, asmongold did you indicate a chat that comments on what the streamer is doing or watching, rather than discussing abstract topics.
Cynical humour: Running joke patterns (om om om) and meme callbacks are common. Humour is used as the primary bonding mechanism, distinct from the political bonding in the hasanabi chat.
Lower echo chamber signal: The presence of do you think and i think as top phrases - without the heavy validation emote layer - suggests slightly more space for differing opinions.
Punitive toward outsiders: Scepticism and dismissal of people perceived as outside gaming culture is common, though internal moderation appears less aggressive than hasanabi's.
Conscientiousness: Low individually - messages are short, reactive, and rarely followed up. At the community level there is some collective conscientiousness around perceived hypocrisy in gaming and media, but it rarely extends to structured or sustained argument.
Openness: Moderate within gaming and entertainment topics; lower on social and political issues where the chat tends toward settled positions. Resistance to change in gaming culture (anti-DEI, anti-"forced" diversity in games) reflects low openness on those specific axes.
Agreeableness: Moderate within the gaming community. Chatters largely agree with each other on gaming culture topics and with Asmongold's reactions. Disagreement with the streamer is more tolerated than in HasanAbi's chat - light teasing and pushback appear without being moderated out.
Authoritarian leaning: Culturally authoritarian in the sense of preferring established norms - particularly around gaming, masculinity, and meritocracy - and resisting what is perceived as externally imposed change. Less interested in enforcing internal conformity than the hasanabi chat, but quick to close ranks against perceived cultural outsiders.
Relationship with the streamer: Chatters mention Asmongold directly at a lower rate - 26,855 messages (4.9% of all messages), roughly half the proportion of HasanAbi's chat. The dominant mode is observational: chatters talk about what Asmongold is doing or reacting to, rather than addressing him. When direct engagement occurs, it ranges from admiring ("sick steering skill asmon", "W @zackrawrr") to playful ("hahaha cooking asmon") to personal - a notable subset share personal stories seeking connection ("I just got the call three hours ago that they might have found cancer"). Approval-seeking exists but takes a less ideological form than in HasanAbi's chat; chatters want Asmongold to notice them or react, not validate a political position.

Linguistic Comparison

Metric	hasanabi	asmongold
Avg message length (chars)	30	24
Avg message length (words)	5	5
Avg word length (chars)	5	4
Avg emojis/emotes per message	0.80	0.39
Vocabulary diversity (TTR)	0.1176	0.1080

TTR (type-token ratio) measures lexical variety: unique words divided by total words in a sample. A score of 1.0 means every word is different; scores approach 0 as repetition increases. Twitch chat scores low by nature - the same emotes, phrases, and reactions recycle constantly - so the comparison between channels is more meaningful than the absolute values.

HasanAbi's chat produces longer messages with slightly more complex vocabulary, consistent with a community whose primary activity is political discussion. Asmongold's chat is shorter and more reactive, consistent with a community watching gaming content and reacting to on-screen events. Neither chat scores high on vocabulary diversity (both TTRs are well below 0.20), reflecting the repetitive, high-volume nature of Twitch chat generally.

Dataset Structure

data/
  hasanabi/          # JSONL, one file per day
  zackrawrr/         # JSONL, one file per day
original-logs/
  hasanabi/          # Raw Chatterino .log files
  zackrawrr/         # Raw Chatterino .log files
analyze_stats.py     # Script that generated the JSONL and stats below

JSONL Schema

Each line in the JSONL files is a JSON object:

{
  "channel": "hasanabi",
  "date": "2026-05-07",
  "timestamp": "21:58:40",
  "type": "message",
  "username": "fake-username",
  "message": "idiots"
}

The type field can be: message, ban, timeout, subscription, watch_streak, or system.

Notes

Usernames have been replaced with unique two-word pseudonyms (BIP39 wordlist, salted hash). Original usernames are not recoverable from the published data.
Chatterino logs include system events (bans, timeouts, subscriptions) alongside chat messages; these are preserved in the JSONL as separate type values.
Messages are capped at 500 characters by Twitch's platform limit, explaining the uniform longest-message length.
Twitch emote detection uses a curated list of global and channel-specific emotes plus an all-caps heuristic. Some false positives and false negatives are expected for channel-specific 7TV emotes.
Both log sets include 14 May 2026 as a partial day.
The Asmongold logs and their corresponding JSONL files are named zackrawrr throughout - this is his Twitch account username, which differs from the channel's public-facing name (asmongold).