I don't know if this new algorithm has fundamental advantages against what Google's currently doing other than it's not yet being targeted by SEO attacks. You can probably create a new algorithm which works significantly better than Google for the current web configuration, but it will be only until everyone decides to "optimize" their site against the new algorithm.
Probably in someday, general computational intelligence could solve this problem by understanding the content itself and use it for ranking in a meaningful way. But if you're just trying to utilize some arbitrary, accidental structure inside the data set for ranking, there always will be a way to exploit. That's the whole point of SEO anyway.
Exactly. Let's say that MarketRank grows in market share to be the dominant search engine. What's to stop someone from then gaming Reddit / HackerNews posts and comments with bots and fake accounts because the operator understands that these websites now have "community verified rank importance"?
They're already gamed or easily gamed. You can buy Reddit accounts with high amounts of karma which are sold to people who are trying to do "astroturf" promotions.
It gobbles up a huge fraction of search results from Google, even things that aren't really image related. Makes finding the original sources for reposted images extremely painful, infuriatingly difficult to avoid without adding -site:pinterest.* to the query, ...
Host a huge number of images that show up in search results and then make them impossible to view without logging in, don't allow right click so you can't save/view image in the normal manner. Basically all the annoying things they can to monetize pictures that people have uploaded.
Instead of using upvotes from sites such as Reddit and HN, why not instead allow the users of the search engine itself upvote and downvote results? This might allow quality results outside the scope of larger online communities to still be highly ranked, and also tailor the results to the audience of the search engine itself.
Because there is no obvious way to translate your own upvotes in to a general search engine. If I want to find info on headphones, there is no way to use my upvotes to show a blog post by a non sponsored non SEO site. At most it could work out I'm interested in headphones and show more of that. But Google already does this.
We have the smartest people working on this problem on both sides. Pretty much every easy / obvious solution has been considered already.
It would know more than that you are interested in headphones. A particular URL that you could upvote or downvote has a ton more information than that - the website itself, the part of the website it is on, the content of the page, other websites it links to. Legitimate sites tend to link with other legitimate sites.
This is all stuff Google already does. Rather than requiring specific upvotes they can generally tell how satisfied with the result you were if you drop off rather than coming back to click more links. Problem is all of these signals are gamed to the extreme.
If that is true then whey do they always suggest Pinterest and Quora at the top when I've been ignoring them for years? If there was a downvote button they would be gone from my results.
Is it that Google actually benefits from having an algorithm that is susceptible to being gamed and therefore they are disincentivized to move away from that?
The more likely but less interesting explanation is simply that once something becomes highly commercially valuable to game, it’s an ongoing cat and mouse situation. Google may or may not see it as a high priority to fix but I doubt it’s a deliberate conspiracy.
Google has several conflicts of interest. The first is that if they provided users with exactly what they are looking for then the users would accept that and not click on any Google search ads.
The second is a lot of the spammy sites are running Google AdSense so it is in Google's interest to send people to those sites even if they aren't what the user is searching for.
> Our naive currency conversion will work exactly like our naive inflation adjustment. We will compare the cost of a similar basket of goods across the different platforms. In this case, our basket of goods will be similar to our inflation calculation, i.e the average of the top 50 highest ranking inflation adjusted websites on a platform.
> We can now calculate the “GDP” of a domain by adding the total value of all webpages produced by a domain.
This is very easy to game by spamming posts with one upvote on any platform, because the formula for calculating score increases strictly positively with the number of links.
You could have low-voted posts count against the total score, but then anyone could easily de-rank a competitor's content by creating low-voted posts about it.
> And so, the oversimplified description of MarketRank is “just add up all the upvotes”.
Wait… the sum of fake internet points is what we’re using to measure pages? I agree that search results could be better but this doesn’t seem to be the way.
Forum upvotes and downvotes have no correlation with quality, though.
As implemented on Twitter and Reddit and HN, they are simply engagement features. In other words, the point of voting is not to surface quality content, it is to make site visitors feel like they’re doing something, and therefore be more likely to return. Karma scores, too, are not about rewarding quality but simply about creating artificial incentives for return visits.
Quality in forums comes mostly from careful human moderation. HN is carefully moderated by dang and others. Likewise, the highest-quality subreddits are those that have strict rules and active moderators that enforce them.
Some of the worst reddits are also the ones with active moderators -- these moderators get under the table money to promote specific content while suppressing the rest.
Agreed. I think the whole market analogy breaks down right out of the gate:
> Each of these communities has an upvote mechanism, which can be interpreted as an indication that the user believes the website is undervalued. They are willing to bid up the price by spending their upvote on it.
It's wholly misleading to talk about "spending" an upvote. Upvotes, unlike money, are not a limited resource. Most of the other problems (bots, spamming, optimizing for low-effort clickbait etc) flow from that.
I don't think ranking based on upvotes, downvotes, or other forms of engagement is necessarily terrible, but it's certainly nothing to do with "market value", and misleading to suggest that it is.
If you want to create a way of ranking websites and call it "SocialRank", or "EngagementRank", that's fine, although see below.
However, users would have to be aware of the severe limitations. You've already highlighted some of the problems with social content: it's often spammy, low effort, clickbait, poorly written, misleading, or outright false.
The quality of the search engine and ranking algorithm is going to be judged by the quality of the results it returns, which I'd imagine will be pretty trashy in this case.
I think you're throwing out an entire idea out based on a small semantic quibble.
Ok, so upvotes on HN and Reddit are unlimited, and therefore not valuable.
Why not limit accounts to X upvotes per day or week? Or only allow upvoting as many times as you've been upvoted yourself? Suddenly upvotes are now scarce.
These aren't impossible problems to solve. HN has implemented tons of "dark" features you don't even realize are here to prevent spamming, and is largely successful at it.
For example, if you post to "New" and send a link to your friends telling them to upvote it, those upvotes will not be counted. Similarly, there is also a wait limit on how soon after visiting HN your upvotes will be counted, without you knowing it. There's a ton of these unwritten tricks like this behind the scenes, and as you can see from the quality level here, they work!
I think ideally the search engine of the future has a combination of user feedback (with HN-style dark patterns behind the scenes to protect integrity), plus some level of human curation.
I'm amazed at how people can simultaneously think Google results are garbage, and yet, also think any attempt to try to solve the problem is futile and worthless. Cynicism is one hell of a drug.
OK, but you're moving the goalposts here. You're not just talking about a new search rank algorithm any more, you're talking about building a new social-web ecosystem to enforce the kind of voting rules needed to make that algorithm work, and somehow getting mass adoption of that new ecosystem.
> I'm amazed at how people can simultaneously think Google results are garbage, and yet, also think any attempt to try to solve the problem is futile and worthless. Cynicism is one hell of a drug.
You're attacking a much broader proposition here than anything I wrote.
I wouldn't say there's goalposts here or winners and losers, we're all throwing out ideas. Issue pops up, humans suggest fix. New issue pops up, new fix needed. This is how all stuff evolves.
I guess I'm more broadly reacting to the sentiment in most comments I'm seeing (which is itself a trope in the dynamics of how upvote-based communities work).
Based off all the recent "Google results now suck" posts that went viral, it seems there's broad sentiment that Google is now a cesspool of SEO'd-to-death affiliate marketing.
But it also seems there's broad sentiment to shut down and discourage new possibilities before they even get attempted, regardless of potential merit.
> Upvotes, unlike money, are not a limited resource.
On some (non-Reddit-derived) sites, they are, although usually you still get a daily allowance and not a persistent budget. There’s also the SO approach where upvotes are infinite but downvotes are limited, that could also work.
If voting didn't exist on HN, don't you think the mods would have had to spend more time curating submissions? The total number of submissions is presumably much larger than the number of reports they receive or the number of suspicious incidents they need to investigate.
Is there a correlation between upvotes and truth or utility?
Sure, upvotes might be fine for recognizing good content or popular opinions. [best resorts in mediterranean], sure, upvotes are as good a signal as anything. Do you think it would be a good signal for queries like [why is there war in ukraine] or [how did covid start]?
This seems incredibly biased and naive. If you want to just search for a tiny fraction of issues relevant to startups and coding, just add you site: filter or use HN search and move on. There's no anti SEO going on here, it's ranking by echo chamber on a metric you hope hasn't been manipulated (that much). Given the the breadth and depth of a fully functioning search engine, this wouldn't cover much at all either.
If the public knows exactly what the algorithm is, they will game it. If they don't know but can approximate its behavior by observation, they will game that.
I also think that lots of upvotes on influential sites will soon turn into improved page rank so don't think that it will make a difference. Google isn't using the original, naive page rank algorithm, they have repeatedly refined it to fight the SEO folks.
A reddit upvote is not a single currency. By this logic a submitted link would get an equal number of upvotes (on average) when submitted to any subreddit. This is not the case. Smaller subreddits have less potential for upvotes.
and so, the oversimplified description of MarketRank is “just add up all the upvotes”. We have some more work to do to make the values accurate, but this is the general idea.
"1,000 FB Post LIKES $15 (On sale for $12)"[1]
"1,000 Facebook Likes $14.5o (On sale for $8.82)" [2]
Facebook likes are much cheaper than reddit or hn upvotes though.
I believe MarketRank is not completely useless, it shouldn’t be used as the only parameter for the ranking system though. Much more a single gear in a complex machine.
why can't we have upvote and downvote on search results, and use that as as measure? i guess a lot of people would downvote spammy sites, not sure about upvoting legitimate ones
Any algorithm is going to get gamed sooner or later. The biggest problem I've with Google ranking is this:
1. Everyone now knows that Google favors long content pieces which covers the topic in depth (what, why, when, etc.)
2. So an army of content marketing firms are writing 2000 word posts for simple topics that can be covered in 200 words.
3. As the user gets lost in the 2000 word article, trying to find what they really need, Google treats this as a positive "lots of time spent on page" signal and rewards this behavior further.
The result is people trying to now write 3000 word articles to "one up" the other already long posts dominating the first few results.
What Google needs is to do is start taking explicit user feedback. If I click on something and it's a major waste of time, I want to be able to tell Google to never show me this result or the whole domain. Why can't I do this? They are happy to hoover all of my data in the name of "personalization", but there's zero way I can personalize the results myself without resorting to insecure 3rd party extensions.
Back in the early 2000s there was a browser toolbar from Google that allowed you to do exactly that. Maybe such a tool today would be too easy to abuse?
Indeed, why does Google not use their full stack? Other competitors have the problem that SEO agencies can flood the feedback with manipulated data. However, with Android, Google knows the identity of enough people that they can identify the genuine feedback.
There is already a search engine for feedback:
Whaleslide.com [1] seems to let users curate content with collections. However, I haven't seen a link to a collection on their results pages.
That is the main value proposition for the new search engine https://kagi.com/
Each result has a menu to block or boost the domain, to remove SEO spam, and pin relevant domains (the online documentation for that framework that always ranks below stackoverflow) to the top. So far I am loving it.
This would be fine on a per user basis, but if your "downvote" informs baseline results, it can also be gamed, with people burying each others' sites by sending malicious negative signals to google about them.
I nearly wept when I looked up some video editing techniques for Shotcut, and saw some from a channel with titles like "Fade in/out Audio in Shotcut in 1min" and the length of the vid was genuinely 1:08 or something.
So many tutorials on Youtube are front-loaded with a long ad advertising the creator's paid courses.
exactly. this post is a kind of: let's make new google! i know how to do it better (while ignoring 1000 other factors outside of my bubble).
Even if Dunning-Kruger effect is debunked, this case shows that people work this way.
This is incredibly naive. Of course this is not immune to SEO but is highly susceptible to SEO. When it is in the financial interests of millions of people with billions of dollars at stake, people will simply tweak the signals going into inputs of your algorithm and "break" your "anti-seo".
What I took away from this article is that the author(s) don't seem to have a feedback cycle outside their bubble.
Hopefully the discussion here might help a bit. I read a lot of valid criticism here that should (in best case) already have been incorporated into the thinking process while researching the idea of a 'Market Rank'.
> An object with a higher market value is probably better
That was the moment the article lost me personally.
Everyone's pretty skeptical here, but I think something like this is the way forward. Not based on Reddit or HN, of course, but based on the same principle, namely, user curation. It's argued that Google was so successful at first with PageRank because at the time it was common for users to have personal web pages that had long lists of curated links to other websites. That signal has gone down the toilet, but why can't we just skip the middleman and get people to rank sites they think are interesting by, you know, upvoting them or something like that. We need collaborative filtering for content aggregation.
Have a quick look at the /newest page here and you’ll see it dominated by people who are already gaming it for conventional SEO, just in case Google starts factoring this in some day.
If you make it Fake Internet Points are worth real money, all you’ll do is fill all these places with SEO people and ruin them.
Rand Fishkin mentions in the opening chapter of his book that HN has an algorithm to detect coordinated upvoting from the same geo ip so you 'need' to have a network of people with different geo ips to game the system.
I knew this was comedy when I saw that their top ranked site was paulgraham.com and they counted hacker news points as almost double Reddit. This is a poorly thought out idea that I just can't get on board with. This also doesn't account for new content that hasn't been ranked on a social aggregator site. I thought it would be about training an AI to recognise good content.
A lot of the Google SEO “secrets” are not so secret. If you have an inquisitive mind - start a new blog with the means of getting your articles ranked on page 1. In 3 months you will know exactly how it works because there is real-time feedback. Google says it doesn’t use site authority, but it clearly does.
People will pay $5,000 (a number i saw once) to get a link from Forbes. Not because they give a shit about Forbes traffic but because Google still thinks that Forbes is some “above them all” content site.
But if you want to see the real dirty stuff people do, buy a subscription to a tool like Ahrefs and see how people manipulate links for SEO purposes. It might just give you gray hairs.
I cannot because I don't have a subscription (it's quite expensive and for my needs I only need to buy it few times a year).
I can tell you this though:
- Brands buy paid guest posts on "established" sites. (Using a stuffed keyword for their money site, service or whatever.)
- They wait until Google indexes their sponsored post.
- They then go ahead and buy links that link to the sponsored post instead of their service / brand site.
- The idea is that they can avoid a penalty in long-term. And Google largely looks past it because the sites doing such tactics (buying hundreds of private network links) are still on page 1 of Google results.
Past a certain point (5,000+ referring domains?) Google isn't going to penalize your site because it can't differentiate between legitimate or spammy links, or even care for matter[0]. Though the internals on the algorithm are very scarce in this regard.
If you do use the tool (Ahrefs) you have full freedom to analyze sites you suspect are doing shitty tactics. And you will find inconsistencies like the one I mentioned above. Needless to say, because Ahrefs is a crawler, it can be blocked and I imagine quite a few blackhat networks do this to avoid further detection.
Lastly, this is a big problem for smaller niches / keywords because all you need is 20-30 links pointing towards an article and Google will assume you have a PhD on that topic.
106 comments
[ 3.2 ms ] story [ 190 ms ] threadWhich would, as you note, instantly become havens for fake votes. They don't get gamed too hard now because they matter so little.
Probably in someday, general computational intelligence could solve this problem by understanding the content itself and use it for ranking in a meaningful way. But if you're just trying to utilize some arbitrary, accidental structure inside the data set for ranking, there always will be a way to exploit. That's the whole point of SEO anyway.
See also the brilliantly named plugin, Unpinterested.
We have the smartest people working on this problem on both sides. Pretty much every easy / obvious solution has been considered already.
Is it that Google actually benefits from having an algorithm that is susceptible to being gamed and therefore they are disincentivized to move away from that?
The second is a lot of the spammy sites are running Google AdSense so it is in Google's interest to send people to those sites even if they aren't what the user is searching for.
> We can now calculate the “GDP” of a domain by adding the total value of all webpages produced by a domain.
This is very easy to game by spamming posts with one upvote on any platform, because the formula for calculating score increases strictly positively with the number of links.
You could have low-voted posts count against the total score, but then anyone could easily de-rank a competitor's content by creating low-voted posts about it.
As implemented on Twitter and Reddit and HN, they are simply engagement features. In other words, the point of voting is not to surface quality content, it is to make site visitors feel like they’re doing something, and therefore be more likely to return. Karma scores, too, are not about rewarding quality but simply about creating artificial incentives for return visits.
Quality in forums comes mostly from careful human moderation. HN is carefully moderated by dang and others. Likewise, the highest-quality subreddits are those that have strict rules and active moderators that enforce them.
> Each of these communities has an upvote mechanism, which can be interpreted as an indication that the user believes the website is undervalued. They are willing to bid up the price by spending their upvote on it.
It's wholly misleading to talk about "spending" an upvote. Upvotes, unlike money, are not a limited resource. Most of the other problems (bots, spamming, optimizing for low-effort clickbait etc) flow from that.
If you want to create a way of ranking websites and call it "SocialRank", or "EngagementRank", that's fine, although see below.
However, users would have to be aware of the severe limitations. You've already highlighted some of the problems with social content: it's often spammy, low effort, clickbait, poorly written, misleading, or outright false.
The quality of the search engine and ranking algorithm is going to be judged by the quality of the results it returns, which I'd imagine will be pretty trashy in this case.
Ok, so upvotes on HN and Reddit are unlimited, and therefore not valuable.
Why not limit accounts to X upvotes per day or week? Or only allow upvoting as many times as you've been upvoted yourself? Suddenly upvotes are now scarce.
These aren't impossible problems to solve. HN has implemented tons of "dark" features you don't even realize are here to prevent spamming, and is largely successful at it.
For example, if you post to "New" and send a link to your friends telling them to upvote it, those upvotes will not be counted. Similarly, there is also a wait limit on how soon after visiting HN your upvotes will be counted, without you knowing it. There's a ton of these unwritten tricks like this behind the scenes, and as you can see from the quality level here, they work!
I think ideally the search engine of the future has a combination of user feedback (with HN-style dark patterns behind the scenes to protect integrity), plus some level of human curation.
I'm amazed at how people can simultaneously think Google results are garbage, and yet, also think any attempt to try to solve the problem is futile and worthless. Cynicism is one hell of a drug.
> I'm amazed at how people can simultaneously think Google results are garbage, and yet, also think any attempt to try to solve the problem is futile and worthless. Cynicism is one hell of a drug.
You're attacking a much broader proposition here than anything I wrote.
I guess I'm more broadly reacting to the sentiment in most comments I'm seeing (which is itself a trope in the dynamics of how upvote-based communities work).
Based off all the recent "Google results now suck" posts that went viral, it seems there's broad sentiment that Google is now a cesspool of SEO'd-to-death affiliate marketing.
But it also seems there's broad sentiment to shut down and discourage new possibilities before they even get attempted, regardless of potential merit.
It seems like an irrational dichotomy.
Any solution is going to have to think deeper than that.
On some (non-Reddit-derived) sites, they are, although usually you still get a daily allowance and not a persistent budget. There’s also the SO approach where upvotes are infinite but downvotes are limited, that could also work.
Is there a correlation between upvotes and truth or utility?
Sure, upvotes might be fine for recognizing good content or popular opinions. [best resorts in mediterranean], sure, upvotes are as good a signal as anything. Do you think it would be a good signal for queries like [why is there war in ukraine] or [how did covid start]?
I also think that lots of upvotes on influential sites will soon turn into improved page rank so don't think that it will make a difference. Google isn't using the original, naive page rank algorithm, they have repeatedly refined it to fight the SEO folks.
"1,000 FB Post LIKES $15 (On sale for $12)"[1]
"1,000 Facebook Likes $14.5o (On sale for $8.82)" [2]
No, this idea isn't going to work.
[1] https://www.fbpostlikes.com/
[2] https://www.instafollowers.co/buy-facebook-likes
I believe MarketRank is not completely useless, it shouldn’t be used as the only parameter for the ranking system though. Much more a single gear in a complex machine.
1. Everyone now knows that Google favors long content pieces which covers the topic in depth (what, why, when, etc.)
2. So an army of content marketing firms are writing 2000 word posts for simple topics that can be covered in 200 words.
3. As the user gets lost in the 2000 word article, trying to find what they really need, Google treats this as a positive "lots of time spent on page" signal and rewards this behavior further.
The result is people trying to now write 3000 word articles to "one up" the other already long posts dominating the first few results.
There is already a search engine for feedback: Whaleslide.com [1] seems to let users curate content with collections. However, I haven't seen a link to a collection on their results pages.
[1] https://whaleslide.com/info/collections
*edit:
also: https://ninfex.com
>There is no crawler, our index is built solely by users submitting interesting and useful links. There is no complex search algorithm.
You can answer this yourself: just think about whether Google is incentivised by the quality of search results, or by something else.
It's like youtube clips of type "10 Best XYZ for beginners" which take 30 minutes while 1 minute is more than enough. Content dilution at its best.
So many tutorials on Youtube are front-loaded with a long ad advertising the creator's paid courses.
Cite please!
This seems like a decent attempt to resolve the the search engine problem. Has flaws yes and is limited to blogs with up votes so not universal.
Perhaps a universal search engine is impossible. Instead each type of content have its own search algorithm
Hopefully the discussion here might help a bit. I read a lot of valid criticism here that should (in best case) already have been incorporated into the thinking process while researching the idea of a 'Market Rank'.
> An object with a higher market value is probably better
That was the moment the article lost me personally.
If you make it Fake Internet Points are worth real money, all you’ll do is fill all these places with SEO people and ruin them.
Please don’t.
So, yeah, it's going on.
https://www.amazon.co.uk/Lost-Founder-Painfully-Honest-Start...
People will pay $5,000 (a number i saw once) to get a link from Forbes. Not because they give a shit about Forbes traffic but because Google still thinks that Forbes is some “above them all” content site.
But if you want to see the real dirty stuff people do, buy a subscription to a tool like Ahrefs and see how people manipulate links for SEO purposes. It might just give you gray hairs.
Can you please provide an example? I'm genuinely curious, I've used these tools but haven't seen the tools to manipulate links.
I can tell you this though:
- Brands buy paid guest posts on "established" sites. (Using a stuffed keyword for their money site, service or whatever.)
- They wait until Google indexes their sponsored post.
- They then go ahead and buy links that link to the sponsored post instead of their service / brand site.
- The idea is that they can avoid a penalty in long-term. And Google largely looks past it because the sites doing such tactics (buying hundreds of private network links) are still on page 1 of Google results.
Past a certain point (5,000+ referring domains?) Google isn't going to penalize your site because it can't differentiate between legitimate or spammy links, or even care for matter[0]. Though the internals on the algorithm are very scarce in this regard.
If you do use the tool (Ahrefs) you have full freedom to analyze sites you suspect are doing shitty tactics. And you will find inconsistencies like the one I mentioned above. Needless to say, because Ahrefs is a crawler, it can be blocked and I imagine quite a few blackhat networks do this to avoid further detection.
Lastly, this is a big problem for smaller niches / keywords because all you need is 20-30 links pointing towards an article and Google will assume you have a PhD on that topic.
[0]: https://twitter.com/KrisHolten/status/1412396082835906563