在本文中，我們將解讀外匯經紀商特意創造的用來吸引客戶的說辭。作者表達了他在外匯平台商工作期間形成的關於商業替代原則的個人觀點。這裡，我們介紹了一系列的實用方法，它們可以幫助你批判性地看待經紀商的活動。本人不涉及那些依照經紀商代理協議運作的IB。我們指的是公司，以自己的名義做交易卻充當清算機構的角色。
“清算機構”: 交易直接轉到銀行同業拆放市場
這種說法是為了掩蓋經紀商與交易者之間的利益沖突。這一說法的提出者試圖強調由於單子是在銀行同業拆放市場中執行，經紀商作為中間方是不會從客戶的損失中獲利的。經紀商的利潤只來源於點差、隔夜利息以及佣金等。
• 根據最流行的模式，流通性提供者基於FIX/FAST協議為經紀商提供報價。比較有名的流通性提供者有巴克萊銀行，摩根大通銀行，瑞士聯合銀行，德意志銀行等等。
• 公司的一個聚合器（比如一個獨立的服務器）負責清算：它從流通性提供者那裡獲得最小的要價以及最高的出價，並為交易者提供良好的交易環境。
• 交易者基於當前價格做出執行交易的請求，經紀商將該請求遞交到大宗經紀機構（夥伴）。
• 大宗經紀機構以報價成交該請求。
這種模式也有變本：一些公司聲稱它們使用公司的賬戶來執行交易請求，以此來維持客戶的匿名性；一些公司直接將客戶ECN賬戶的交易請求遞交到同業銀行。就複雜性和多階段性（多至6步的數據傳輸）而言，這些方案是類似的。在揭穿這些說辭之前，讓我們列舉出幾點使新公司維持競争力假象的原因。
• 與專業交易者不同的是，交易新手除了盈利之外還有一系列別的交易動機。這些動機無一例外地阻礙成功交易，而且很難去除。其中一個動機是來源於金融領域的自負：一些人天真地認為每一筆交易請求都在巴克萊或者摩根銀行處理。
• 交易新手對數據視而不見並且高估了利益沖突。費城金融，美國的一家投資公司，它對美國零售經紀商公佈的數據做了一項獨立的研究。經紀商是否有任何理由為投資者人為制造困難？存款預期虧損是否由過程心理決定？賭博心理的興奮性以及缺少資金管理使得外匯經紀商可以像賭場一樣獲得可觀收益。你真的認為賭場會不顧及名聲地去幹預輪盤賭或者標識撲克牌？
• 交易者不理解保證金交易的原則。銀行間貨幣操作的比率是1：1。缺乏流動性的是交易者而不是銀行。槓桿交易意味著經紀商需要提供自己的資金來傳達一筆交易。如果槓桿是1：500，那麼經紀商就需要50萬美金來開設1000美金的倉位。如果將客戶的交易請求數量計算在内的話，經紀商需要有最大的投資銀行所擁有的資金量。
• 最後，失敗的交易者尋找他交易策略之外的失敗原因——稱作“受害者心理”。承認自己的錯誤對於他們來說相當的困難。但這不意味著你不應該批判性地看到經紀商的服務，同時，也不說明不應該誇大經紀商的缺點。
有兩種方法可以打破經紀商的神話：技術手段以及司法手段。兩種方法都可以得出：同業機構模式實際上是不適合外匯經紀商的競争環境。
技術沖突
你是否曾經想過，為什麼對沖基金會將它們的服務器直接的放置在證券交易所大樓？市場直接準入（DMA）技術是唯一可以避免跨平台延遲、降低訂單執行時間至0.5~3毫秒的技術。這樣一來，公司就可以使用銀行或者證券公司的設備並且可以擁有自己的操作環境、決定報價。如果對沖基金使用算法交易，那麼它會經常地更新自己的軟件以及下載新的程序模塊。意思是，如果一個公司的交易服務器與外匯交易所或者是別的證券公司的服務器位於同一個房間，那麼它可以在1毫秒内獲得報價。否則，它就只能在延遲10-100毫秒後得到以太網或者延遲500-600毫秒得到衛星連接的報價。
代理服務器以及信號中斷服務器的使用導致了報價的更長時間的滞後性。每一個代理服務器都會收到無數的交易請求並按照順序執行，這在商業的DMA模式中是不存在的。你曾經是否想過，一個訂單在收到報價與執行之間會有多長時間的延遲？為了測試你的電腦連接經紀商的服務器，你可以使用一些簡單易用的程序或訪問http://ping.eu/ping/，輸入主機服務地址器或者IP。
“Avg”字段顯示了傳遞一個信號（交易）到服務器並執行它所需要的平均時間。結果比DMA執行模式的效果差了300倍。你可以使用這個程序測試與別的經紀商（聲稱自己是中間銀行機構）服務器之間的速度，測試結果同樣是100-1000毫秒的延遲。這個測試得出的主要結論是：目前的中間銀行執行一個請求時間是500 – 1000毫秒。因此，代表流通性提供者執行交易甚至主要經紀商都是不可能的。
假設服務器執行訂單有500毫秒的延遲，這種情況下，在1.12770點位賣出在銀行執行之前將會導致35次重複報價。我們現在很不幸運，並且價格沒有在3秒内回到該價位，那麼經紀商是如何反映的？
1. 它向交易者重複報價（這種情況下是5~6次）；
2. 它自己承擔風險開倉，並等待價格回歸該區間。
在第一種情況下，服務質量大幅下降。如果你準備在市場平靜時期重複6次報價，那麼這就是你要尋找的模式。第二種情況，經紀商脫離機構範疇並充當了做市商，因此，它不再實際上地傳遞頭寸。為了訂單的最優執行，經紀商現有狀態下能承擔的最大延遲是30~1000毫秒。
FX brokerage: view from inside
In this article we try to dispel a myth, intentionally created by FX brokers to attract clients. The author expresses his personal opinion, formed when working for a FX broker with alternate business principles. We express a number of practical methods that allow you to examine critically broker's activity and will probably let you dispel the myth yourself. The article doesn't concern introducing brokers (IB) that openly operate conforming to agency agreement. We refer to companies, making deals on their behalf but positioning themselves as clearing agents.
“Clearing agent”: deals are transferred directly to interbank market
This myth is aimed to cloak the broker-client interest conflict. Authors of this myth underscore that a broker-agent doesn't stand to benefit from trader's loss because the deal is executed in interbank market with the broker being an intermediary. The only broker's profits are spread and swaps, commissions etc.
• According to the most popular model (figure below), the company-agent receives a quotes stream from several liquidity providers using FIX/FAST protocol. Further on, respected establishments are listed: Barclays, JP Morgan, UBS, Deutsche Bank etc.;
• A company's aggregator (i. e. a separate server) carries out clearing: it forms minimum Ask and maximum Bid, following providers' streams, and creates optimal conditions for clients;
• A client makes a request to execute a deal at the current price; the request is transferred by the broker-agent to a prime broker (partner);
• The prime broker executes the request at the quoted price.
The model has variations: some companies assert that they execute a request from the company's account, preserving a client's anonymity. Some - that every trader's request is transferred directly from his or her ECN account to interbank. These schemes are alike in relation to their complexity and multiple stages: up to six steps of data transferring. Before busting the myth, let's go on and name the reasons that make newcomer companies sustain the illusion to remain competitive.
• A beginning trader, unlike a professional one, has a number of other motivators except making money. These motivations impede successful trading and don't let be impartial, but it's hard to throw them off. One of the motivations is the sense of self-importance, engendered by belonging to financial sector. Someone really likes to think that every request is processed by Barclays and JP Morgan;
• A beginning trader turns a blind eye to statistics and overestimates conflict of interests. American investment company, Philadelphia Financial, undertook an independent study based on data, revealed by the US retail brokers. The chart below shows the probability of a complete deposit loss. This brings up the question: is there any sense for a broker to create artificial difficulties to an investor, if deposit loss expectancy is determined by process psychology? Gambling excitement and lack of risk management makes an FX broker almost as cost-effective as a casino is. Do you really think that a casino tampers roulette or marks cards, mindlessly risking its reputation? If so, take a look at the statistics of putting stake on Zero;
• A trader doesn't understand the principles of marginal trading. Interbank currency operations have the 1:1 leverage. Banks do not lack liquidity. A client does. That is why trading with leverage means that a broker has to provide its own funds to transmit a position. If the leverage is 1:500, a broker has to deliver $0.5 million own funds to open a $1000 order. Taking into consideration the number of clients' requests, a broker has to possess that much liquidity as biggest investment bank have;
• Finally, a losing trader feels like looking for an external reason beyond his or her strategy – so called “victim mentality”. It is unbearably hard to admit mistakes of your own. But it doesn't imply that you should not critically analyze broker's services, yet, it doesn't mean you are supposed to overestimate broker's disadvantages.
There are two methods of busting this myth: a technological one and a juridical one. Both methods lead to the following conclusion: interbank agency model is practically impossible within the competitive environment of FX brokers.
Technology conflicts
Have you ever wondered why hedge funds place their servers directly inside a stock exchange building? The Direct Market Access (DMA) technology is the only way to reduce request processing time to 0.5-3 ms and avoid additional cross-network delay in order execution. In this case a company uses bank's or stock's equipment and has the opportunity to imbed its own program environment, determining quote stream. If a hedge fund uses algorithmic trading, it occasionally upgrades software and downloads new program modules. That means that if a company's trade server is located within the server room of MICEX or another stock venue, a typical quote resolution available will be 1 ms. Removed access gives additional signal delay of 10-100 ms for Ethernet or 500-600 ms for satellite link.
Proxy servers and signal repeaters usage results in longer delays. Every proxy server receives millions requests and lines them up in a queue, creating a time lag, which is a priori absent in DMA business model. Have you ever thought about how long the delay is between receiving a quote and executing an order (except for delays caused by user)?
To test lag in transmission from your PC to a broker's server you may take advantage of several easy-to-use programs or services ashttp://ping.eu/ping/. Enter an IP or a host server address in the field. Below is the ping of “agent” trading proxy-server when connecting from my home PC (100Mbps Ethernet).
The “Avg” field shows overall time, needed to transmit a signal (an order) to server and confirm it, i. e. to make one trade. The result is 300 times worse than with DMA. You may repeat the experiment with any other server of a broker, describing itself as an “interbank agent” and get the same delay in a range of 100-1000 ms. The basic conclusion from the test is as follows: to execute a request at the current interbank price it has to remain the same for 500 – 1000 ms. Otherwise execution on behalf of a liquidity provider or even of a prime broker becomes impossible. Let us consider the EUR/USD BID and ASK chart by the biggest liquidity provider - Ducascopy bank. It is 17:00 (MSK), 02.14.2015. Market is relatively calm: there has been no important macroeconomic news or indicators for one hour.
A 1000-ms-wide window is chosen at random. Inside that window 44 ticks are located. If you move it 1 step forward, the tick number will rise to 50. Let us assume that a trade server executes orders with a 500 ms delay. In this case a request to sell at 1.12770 will lead to 35 requotes, before a bank confirms it. Currently we are unlucky and the price hasn't returned to the stated level in 3 seconds. What is a broker's reaction?
1. It sends a requotes to a client (5-6 requotes in this case);
2. It opens a position at its own risk, waiting for the price to be back into the range.
In the first case, services quality falls sufficiently. In the second case, broker goes beyond agent model boundaries and becomes a market maker, i.e. doesn't really transmit position anymore. For best requests execution the maximum delay a broker can afford is 30 ms or 1000 ms/N in the current market stance. If you are ready for 6 times requotes in the calm market, that's the model you are looking for!
本文翻譯由兄弟財經提供
文章來源：http://www.fxstreet.com/education/forex-basics/fx-brokerage-view-from-inside/2015/03/10/