Pyth Network’s Vision for the Future of Financial Market Data
In the financial ecosystem, where the role of data is pivotal, Pyth Network emerges as a beacon of innovation. Guided by the acuities of Mike Cahill and Jayant Krishnamurthy, CEO and CTO at Douro Labs respectively, it aims to revolutionize the perception and utilization of data in the financial sector.
We had the pleasure to speak with Mike and Jayant. Here’s what they said:
Can you explain what you do to your grandma?
There’s a bunch of people building a new financial system powered by a technology called “blockchain.” This system will be fairer and more efficient than the current one. But this system needs data from the outside world. That’s where Pyth Network comes in. The mission of the Pyth Network is to bring all of the world’s finances to the blockchain.
Now, Pyth isn’t the first to build an oracle. But it looked at what the oracle landscape was like and addressed some longstanding issues. Past oracle solutions had limitations on the type of data they could access; they relied on free data, updated slowly; or they weren’t available on all blockchains. What makes Pyth Network special is it gets the data straight from the source (financial institutions), sends it faster than competitors and is accessible on all of the 30 blockchains the network covers.
Products and Services
With over 80 first-party publishers contributing to Pyth, how do you ensure the integrity and accuracy of the data? What measures are in place to prevent manipulation?
The institutions that contribute data to the Pyth Network are known as data providers or “publishers”.
Pyth data providers are typically well-established, credible institutions that possess a wealth of high-quality data, including global exchanges, market makers, and trading firms. Some of the most recognizable names include Cboe Global Markets, Jane Street, Optiver, Binance, OKX, QCP Capital, Two Sigma, Wintermute and CMS.
All of these data providers are first-party sources: they create and therefore own the price data they contribute, as they are either trading venues receiving orders (price at which traders intend to trade) or are traders themselves (and executing trades at specified prices).
First-party data means assurance of data quality and network security. The contributions by all data providers for any Pyth data feed mean that individual data sources can be held accountable for the quality of their inputs. Furthermore, the reputations of these data providers, and the knowledge that a malicious attack on their end would have a detrimental impact on their business as a whole. This is a powerful and additional deterrent layer against traditional oracle attack vectors.
The second layer is the on-chain aggregation methodology. For every 400ms or “Solana slot”, the Pyth oracle runs an on-chain, transparent aggregation algorithm on the prices and confidence intervals published by each data provider for a symbol on the previous slot. This aggregation produces a single aggregate price and confidence.
The aggregation mechanism is designed to have three main properties:
to be resistant to manipulation - for example, when a data provider publishes a price too far away from the other components in an attempt to manipulate the aggregate
it weighs data sources with different levels of accuracy - for example, when a data provider is more confident about the price than another
to take into account variations between publishers - because assets trade at different prices across different venues.
So Pyth does not publish data itself, and it uses an aggregation mechanism to protect price feeds.
There’s another component to ensuring the quality of prices - and that is the conformance testing process.
Before Pyth announces a new price feed and before a new data provider is allowed to contribute to a price feed, there is a comprehensive conformance testing process. For new data providers, the confidence interval logic is checked over to make sure it is set up correctly and that the provider is not publishing a price with a confidence interval that is too wide or too tight.
A reliability model is run to detect any potential issues to make sure there are no providers going down at the same time, or that prices are not too correlated.
A synthetic aggregate price from historical data is also computed, to assess the quality of what the aggregate would look like.
Pyth offers three different data products: Solana Price Feeds, Pythnet Price Feeds, and Benchmarks. Can you explain the unique value proposition of each and how they cater to different applications?
Pythnet Price Feeds is a technical term for real-time Pyth Price Feeds, which are available on more than 30 blockchains, including Ethereum, Arbitrum, Optimism, zkSync Era, BNB Chain…
Pyth currently has more than 350 Price Feeds that cover cryptocurrencies, FX, metals, ETF’s and US equities. Because price feeds are subsecond, and cover many assets and blockchains, DeFi applications can now offer financial products like perps that were previously not possible with other oracle solutions.
“Pythnet” is the name of an application-specific chain (appchain). The Pythnet appchain is based on Solana’s technology, but it’s actually a separate network from Solana mainnet-beta.
As Pyth Network grew it became clear it needed its own application-specific chain to avoid congestion and increase reliability. With Pythnet DeFi applications can get reliable, low-latency prices without the risk of congestion.
Solana Price Feeds is a technical term for a push oracle design that still runs on Solana mainnet-beta for Solana users only. Pyth Network started off on Solana, with the Pyth protocol and aggregation program being on Solana. One day, Solana users will also be using Pythnet Price Feeds. It’s best to call this product “Pyth Price Feeds” to avoid confusion.
Last is the Pyth Benchmarks. Benchmarks are a broad class of standards used in the financial world to steer market participants’ decisions and arbitrate payouts.
Pyth Benchmarks provide access to historical Pyth prices for both on- and off-chain applications. Users can access this data via the new Benchmarks page on the website, which lets users search the dataset of historical asset prices. These prices already meet all the requirements of a good benchmark: All of the source data is provided by reputable first-party publishers, and the data is robustly and transparently aggregated on the Pythnet blockchain, and anyone can verify the computation. Pyth Benchmarks offers historical and accurate data for options settlement for players like Ribbon Finance.
What challenges did you face in integrating real-time prices across 30+ blockchain ecosystems, and how did you overcome them?
The main challenge is scalability: the oracle needs to be architected to scale to many different blockchains. Previous oracle designs were not scalable, as the oracle had to pay transaction costs for every feed on every chain. Consequently, they can’t support lots of new chains, because it would bankrupt them.
Pyth solves this problem with its pull architecture. The oracle generates a stream of signed price updates that anyone can permissionlessly pull onto any supported blockchain. This architecture eliminates transaction costs and is the core innovation that enables Pyth to support 30+ blockchains.
There are of course a number of other technical challenges in supporting 30+ blockchains. Every blockchain has its own unique programming environment and quirks, so every new chain has a learning curve. However, thanks to the pull architecture, there’s a relatively small amount of work per chain, so this learning process doesn’t take too long.
The Founders' Journey
What inspired both of you to create Pyth? Can you share the origin story and how the idea evolved into what it is today?
Before Pyth, Mike was researching existing oracle solutions and how they work. He realized that the largest legacy oracles were essentially scraping free data and that this model is not sustainable. Anyone who’s been around financial markets knows that market data is valuable. The big traditional finance exchanges like NASDAQ make something like 20% of their revenue (that’s about $6.5B today) from selling market data. If the crypto markets play out the same way – and we’ve already seen that Coinbase started charging for data last year – then collecting data from public sources and aggregators isn’t going to work. The market data will be sold with a license that prevents redistribution, and then anyone copying that data on-chain (like an oracle node) will get sued.
That insight led to the idea of creating a first-party financial oracle, where the owners of the data directly contribute to the network. We started talking to major market participants and exchanges, and found a lot of interest in this idea. These market players started joining the network to contribute their own proprietary, valuable data. Pyth Network now has over 80 first-party data providers contributing to the network, and these are all major firms like Two Sigma, Virtu, Binance, Cboe.
As founders, what have been some of the biggest obstacles you've faced in building Pyth, and how have you navigated them?
Pyth started building on Solana – because it needed a fast blockchain for financial data, because speed is important – and launched there in 2021. It quickly became the dominant oracle on Solana, securing something like 90% of the value. That was the first big success.
From there, Pyth Network started expanding to other blockchains. This was always on the roadmap, but it was definitely a challenge, especially since very few people have built protocols that work across multiple blockchains.
Today Pyth is deployed on +30 blockchains. Each deployment comes with its own set of challenges, from navigating the language, tooling, libraries, and other nuances. Our team has had to grow to become experts in each chain and then start from scratch for the next one, which is incredibly humbling.
Staying focused during the bear market has been incredibly important as well. There are some elements where this is actually a benefit as the tourists have mostly left. This means we’re mainly dealing with the really talented developer community who are going to bring the next wave of innovation. However, it’s important to avoid distractions and hold our thesis conviction on blockchains high. As an infrastructure provider, Pyth has had incredible adoption during the bear, with 4 new application integrations per week in 2023. This keeps us motivated.
Future Vision and Industry Insights
How do you see the future of financial market data in the blockchain space? What innovations are you working on to stay ahead of the curve?
Cryptocurrency markets are unusual in that data such as order book data is free and easy to access. In contrast, the traditional financial markets require costly subscriptions for this kind of data.
This will not last forever, however. Coinbase, for example, began charging for market data last. Binance doesn't charge anything. The traditional exchanges didn't charge for market data until about 15 years ago, and then it became a large portion of their revenue, and they've segmented and itemized it, where it actually feels productized, as opposed to, “we’re just going to extract more money because we think we can”. That will inevitably happen [to] crypto data as well.
Pyth Network is staying ahead of the curve by bringing the world’s largest network of first-party data owners to the blockchain.
All of the Pyth data providers understood that anyone betting on Web3 and DeFi in particular to grow, needs to first solve the oracle problem. They also understand that the solutions before Pyth were incomplete and unlikely to scale to support multiple assets. For these data providers, Pyth Network fulfills these prerequisites to enable the Web3 use cases that are the most interesting for them.
What gaps still exist in the market data space, and how does Pyth plan to fill them? Can you share any upcoming features or partnerships?
The symbols that are currently on the Pyth Network are a combination of those with the highest demand in addition to some that are representative of things that Pyth is uniquely able to publish, such as real-time US equity data. We expect the network to grow based on the demand and for this to be controlled by the community.
Pyth is focused on first-party data providers. They have the most control over the quality of the data they submit. 3rd party data aggregators are not great. Most of the good ones have the appropriate licenses, but there are some that have limitations with regard to their legal right to redistribute. They also have limitations to adding functionality like confidence intervals or having full control over their publishing speed.
Advice and Reflection
What advice would you give to new entrepreneurs entering the blockchain and crypto space? What lessons have you learned from your journey with Pyth?
Innovating within the blockchain space has a move-fast theme, but it’s paired with don’t break things. When designing or developing in Web2, the consequences of bad code are typically bad UX, bugs, or relatively low-stakes glitches. In Web3, the stakes are incomparable, as it tends to be an adversarial environment where weakness can be exploited for potentially huge sums of money. At Douro, we take security incredibly seriously, while still shipping as fast as possible. Web3 entrepreneurs should not underestimate the effects of bugs.
On the more positive side, Web3 builders are some of the nicest people you could ever imagine working with. The space is still early enough to feel positive-sum, and most builders are hoping they and their competitors both succeed. Now is a great time to enter into the space.
How has Pyth impacted the broader crypto and financial industry? Can you share any success stories or significant milestones?
48% of DEX volume uses an oracle.
The Synthetix Council made the landmark decision last December to transition to off-chain price oracles with on-chain validity verification for their Perps V2, thereby adopting Pyth Network as their primary oracle and facilitating more than $22B in trading since its launch. Their community-written performance report said it best: “Without the need to thwart front runners through cost-prohibitive pricing, fees have fallen dramatically to the point that Synthetix perps are among the most competitively priced in the DeFi space.” The Synthetix ecosystem took notice quickly, and now Kwenta, Lyra Finance, Polynomial and Decentrex are integrated with Pyth Price Feeds.
Was Ryze a good investor to team up with?
Building a successful business in the blockchain and Web3 space is no small feat. It requires not only innovative ideas but also strong financial backing and strategic guidance. As contributors to a blockchain oracle network, we understand the importance of choosing the right partners on this journey.
It’s been a pleasure having Ryze Labs in Pyth Network’s corner. Their strategic insights and their ability to connect Pyth Network to so many talented teams across such a diverse industry have been a huge value add. Above all else, they “get it”. They have a specific vision for Web3 that is thoughtful, imaginative, and logical. They’re one of the few teams who understand just how important DeFi infrastructure is.
There’s also a cultural fit. We once asked Matty to do deadlifts for a surprise video, and he delivered. Great stuff.